Home

MIMOSA user's manual (Draft version 1.4.0) - UR Green

1. StupidModel 1 BugPopulation lt lt Scripting gt gt nbBug Integer StupidModel1 Bug position lt lt Scripting gt gt tite observer for i in range nbLine for j in range nbCol addPort port cell i nbLine j StupidModel1 Cell parameters pair x i pair y j sendProbe space nbLine nbCol sendLogical observer spaceSize nbLine nbCol e a bug has in charge to position itself randomly at initialization it could also be done by the position and to move at each time step For initial ization it just asks to the position to do itf sendLogical position randomPlace no For moving it is a combination of three functions Aint to say when to perform the next move Aert to ask the position to move the bug and dint to perform the internal transition Given that the only action to do is to ask the position to move the bug the latter function does nothing So we define Aert sendExternal position randomMove no 4 assuming that position shall perform the random move within the given distance 4 and Aint sendInternal 1 moveRandom e a cell is doing strictly nothing 4notice that it is a logical influence because it is just used to initialize the simulation model it is not properly speaking an action CHAPTER 7 SOME EXAMPLES 70 e the position is doing all the work by managing the position of each bug and by p
2. The buttons of the model editor The creation dialog for an individual The individual graphical form The individual editor with the attribute and the initialization A concrete model with an output specification The creation dialog for an output LIST OF FIGURES 6 9 The output graphical form 6 10 The output editor with an attribute panel 6 11 The creation dialog for an output edge 6 12 The control panel toolbar 6 13 The control panel view 6 14 The creation dialog for an output view 6 15 The creation dialog for a parameter editor 7 1 The conceptual model for a kicked and observed rolling ball 7 2 The conceptual model for a rolling ball with the attribute panel 7 3 The rolling ball category with the relations panel 7 4 Definition of an arc from an existing relation definition 7 5 The rolling ball category with the probes panel 7 6 The rolling ball category with the initialize panel 7 7 The rolling ball category with the external transition panel 7 8 The concrete model as an instance of the conceptual model 7 9 The edition dialog for an individual 7 10 The view on the rolling ball state 7 11 The stupid conc
3. 8 3 The entity inspector window 8 4 The graph of the simulated model Chapter 1 Introduction Mimosa is an extensible modeling and simulation platform 9 It is aiming at supporting the whole modeling and simulation process from the conceptual model up to the running simulations The modeling process is assumed to be constituted iteratively of the following stages The conceptual modeling stage it consists in elaborating the ontology of the domain as a set of categories their attributes and their relationships either taxonomic or semantical In short this stage is about defining the vocabulary used to describe concrete situations the words to say it The dynamical modeling stage in order to describe the dynamics of the categories defined in the first phase one must decide on the choice of paradigm differential equations straight scripting agent based etc for each category The paradigm is described using a built in meta ontology Given the choice of dynamical paradigm one must specify the possible states and state changes according to the chosen paradigm Additionaly one can define how to initialize the states and how to observe it the concrete modeling stage the previously described stages define the vo cabulary in which the concrete model s can be described as a set of indi viduals linked to each other and with given attribute values the simulation specification st
4. e the down part is used to create a new output with a drop down menu from which to select the kind of output one wants to create CHAPTER 6 THE CONCRETE MODEL EDITOR 53 ANA Concrete model StupidModel3 StupidModei3model xml DefaultModel Concrete model Control panel jel3 AAA AA a Conceptual model DefaultOntology a foal Bias S s seed 5612653141 population space Discrete2DSpaceView outl observer StupidModel3 BugPopulation nbBug 20 nbLine 20 nbCol 20 Figure 6 7 concrete model with an output specification AA Existing Discrete2DSpaceView out2 em Qutpue ProbeView 4 New Existing Cancel Figure 6 8 The creation dialog for an output Figure 6 9 The output graphical form CHAPTER 6 THE CONCRETE MODEL EDITOR 54 Output Discrete2DSpaceView a Cell Object Layer INF 0 2 0 Color INF 5104 0 Color INF wie 0 6 1 0 Color food INF o s 1 0 Color ELSE 0 0 Color More conditions Less conditions More layers Less layers Ok Cancel Figure 6 10 The output editor with an attribute panel An ellipse is drawn as illustrated in the figure 6 9 with a name which com posed of the kind of chosen output and an automatically generated name be tween parenthesis to uniquely identify this output
5. package example import mimosa state DefaultState public class MyState extends DefaultState public void setParameter String name Object value throws EntityException public void dolnstancelnitialize StateSpecification stateSpecification throws EntityException The two provided methods are called when an entity is created setParameter for possibly storing the parameter values if necessary and dolnstancelnitialize for building the structure of the state The latter method is parameterized by an externally setup specification which is entirely dependent on the way the state is defined We shall not describe further this later possibility but see the programmer s manual for more details For defining how to initialize the state each time the simulation is run the resulting file could be like this one package example import mimosa init AbstractStatelnitializer import mimosa scheduler EntityException import mimosa state State public class MyStatelnit extends AbstractStatelnitializer public void initialize State state throws EntityException The method initialize is called each time the simulation is run The state is given as a parameter because the role of this class is to initialize the state If the way to intialize the state is parameterized externally the variable initializerSpecification contains the related specification Once again this structure is completely pro grammer dependent and shall
6. 1 sendLogical position randomPlace no sendLogical population size no size For moving and eating it is again the combination of three functions Aint to say when to perform the next actions to ask the position to move the bug and the cell to get food and int to perform the internal transition Given that the only actions are requests to others the latter function does nothing as in function one So we define Aert sendExternal position randomMove no 4 sendExternal cell consume maxConsumptionRate and Aint sendInternal 1 moveRandomAndEat Finally the cell shall inform the bug of the eaten quantity by a logical influence to be handled by dog for inf in getAllInfluences if type inf consumed consumed contentOf inf size size consumed sendLogical population size no size The bug grows according to the actually received food The influence could have been an external influence but because an external influence is only issued at the next transition the food would have been consumed only at the next step e acell has now a certain quantity of food which is growing over time Ad ditionally it has to react to food consumption requests The initialization becomes as follows availableFood 0 maxFoodProduction 0 1 rand newRandom 6538672547 sendLogical space food x y availableFood Notice that the gro
7. INF Tao y En Color Shade color Ellipse 5 width 5 0 Theight 5 0 INF 4 0 Color v Shade color Ellipse width 5 0 height 5 0 INF Pe 6 0 Color M Shade color Ellipse width 5 0 height 5 0 bug ale f gt INF 80 Color JM Y Shade color Ellipse width 5 0 height 5 0 ELSE 33 oc Color v C Shade color Ellipse width 5 0 height 5 0 More conditions Less conditions More layers Less layers More types Less types Ok Cancel Figure 7 17 Defining the visualization of the bug size single visualization in addition to the possibility to send probes to a variety of observers 7 3 3 Stupid model 3 The cells contain food of which quantity is growing over time The bugs are eating what is on the current cell and grow accordingly Defining the conceptual model The resulting conceptual model is shown in the figure 7 19 The only modifi cation is the reciprocal relation between the bug and the cell because of their interactions Because of the important induced changes we shall describe again in turn each of the category e the bug population does not incur any change because everything is in place for monitoring the bug size changes e the space has to work the same way as the bug population for monitoring the cell food availability Consequently a link has to be created from each cell to the space for i in range nbLine for j in range nbCol
8. cell port cell i nbCol j addPort cell StupidModel13 Cel1 parameters pair x i pair y j linkPort portRef cell space self lt HERE sendProbe cellLayers food sendLogical observer spaceSize nbLine nbCol As for the bug population the space is informed of the food availability changes through logical influences to be handled by the djog function for inf in getAllInfluences if type inf food CHAPTER 7 SOME EXAMPLES M Show grid 1 Spread objects Select cell layer Select object layer ea 4 13 a el Figure 7 18 The simulation model 2 control panel population StupidModel3 BugPopulation lt lt Scripting gt gt sg imager LtupidModel3 BugPosition lt lt Scripting gt gt gt mr Six _seed L Pace 115p bug 1 StupidModel3 Space StupidModel3 Cell space lt lt Scripting gt gt lt lt Scripting gt gt 1 ai nbLine Integer nbCol Integer Figure 7 19 The stupid conceptual model 3 CHAPTER 7 SOME EXAMPLES 77 x y food contentOf inf sendProbe cellState x y food food e a bug has now in charge to define its size and maximum consumption rate at initialization size 1 maxConsumptionRate
9. lt left gt lt right gt A list lt elt gt lt elt2 gt lt elt gt is nothing but lt elt gt APPENDIX A INTRODUCTION TO SCHEME 99 lt elt2 gt lt elt gt following functions O where is the empty list We have the pair lt exp gt null lt exp gt list lt exp gt car lt exp gt cdr lt exp gt set car lt pair gt lt obj gt set cdr lt pair gt lt obj gt list lt obj1 gt lt obj gt length lt list gt reverse lt list gt list tail lt list gt lt k gt list ref lt list gt lt k gt append lt listi gt memq lt object gt lt list gt memv lt object gt lt list gt lt list gt tests if dotted pair tests if empty list tests if empty list or dotted pair left of dotted pair or first element of list right of dotted pair or rest of list modifies left of dotted pair modifies right of dotted pair creates a list length of a list reverse of a list the k th rest of a list the k th element of a list append of lists member using eq member using eqv member lt object gt lt list gt member using equal An additional structure is the so called a list which is a list of pairs whose car is considered as a key and the cdr as the associated value The related functions are assq lt object gt lt list gt has key using eq assv lt object gt lt list gt has key using eqv assoc lt object gt lt list g
10. Project browser opens a window for managing and editing the projects as a set of conceptual concrete and simulation models Conceptual model browser opens a window for editing concep tual models Concrete model browser opens a window for editing concrete models as instances of conceptual models CHAPTER 2 RUNNING MIMOSA 11 Scheduler opens a scheduler control window for running the sim ulation models Open loads the content of a file depending of the selected window or item if in the project browser The file must contain an appropriate XML representation The kind of content which can be loaded de pends on the active window If it is a conceptual model editor only a saved conceptual model can be loaded If it is a concrete model editor only a saved concrete model can be loaded In the last case be sure that the conceptual models used by the concrete model have been loaded beforehand Finally if it is a scheduler window only the XML files especially generated from the concrete model editor for this purpose can be loaded Save saves the model currently edited in the active window in the asso ciated file the last file it was saved to If it was never saved before a file chooser dialog opens Save as saves the model currently edited in the active window in a file to specify regardless of the last save or open Save as image saves the displayed graph if any as a picture in a number of proposed formats
11. Running the simulation model The figure 7 22 shows the resulting visualization while running the simulation The cell colors are shown according to the quantity of available food Once again it illustrates the possibility to combine multiple sources of data within a single visualization 7 3 4 Stupid model 4 This version does not add any new behavior but provides the possibility to monitor individually the cells and bugs through the grid interface Defining the conceptual model The conceptual model structure does not change and remains the one shown in the figure 7 19 The grid visualizer as described in 6 3 2 already provides the possibility to probe the cells and objects which are visualized However for it to work e the BugPopulation and the Space must provide the ports to access the entities to monitor by using the probes cellEntity lt x gt lt y gt lt port gt and objectEntity lt type gt lt id gt lt port gt as described in 6 3 2 e each cell and bug must also send probe to advertise their state changes Accordingly CHAPTER 7 SOME EXAMPLES 80 000 i Output Discrete2DSpaceView r Layer Figure 7 21 Defining the visualization of the food availability M Show grid 1 Spread objects Figure 7 22 The simulation model 3 control panel CHAPTER 7 SOME EXAMPLES 81 e the bug population does inform the probe observer for each of its bugs for
12. as short int long float double in Java existing in a programming language for computational purposes The corresponding numerical value only exists relative to a unit which is considered as a reference with respect to quantities C 2 4 The units Bibliography 1 2 3 4 5 6 7 8 9 10 11 12 13 14 http bergel eu athena http herzberg ca sandia gov jess http scheme sourceforge net jscheme main html http www alice unibo it 8080 tuprolog http www beanshell org http www gnu org software kawa http www jython org Jacques Ferber and Jean Pierre M ller Influences and reaction a model of situated multiagent systems In Mario Tokoro editor Proceedings of 2nd International Conference on Multi Agent Systems pages 72 79 Kyoto Japan December 1996 AAAI Jean Pierre M ller The mimosa generic modeling and simulatiion plat form the case of multi agent systems In Herder Coelho and Bernard Espinasse editors 5th Workshop on Agent Based Simulation pages 77 86 Lisbon Portugal May 2004 SCS Jean Pierre M ller Mimosa using ontologies for modelling and simulation In Proceedings of Informatik 2007 Lecture Notes in Informatik September 2007 Jean Pierre M ller Towards a formal semantics of event based multiagent simulations In Proceedings of the Multi Agent Based Simulation Workshop Estoril Portugal May 2008 Jean Pierr
13. equivalent to Aint together e public void doGetLogical State self throws EntityException equiv alent to Alog e public void doGetStructural State self throws EntityException equivalent to Astr If something is going wrong just throw an EntityException with the entity and a message as parameters The exception will be taken into account by the architecture in an appropriate way Do not forget to catch any possible exception and raise an EntityException accordingly for securing the model execution Because they are predefined for doing nothing you can only define the methods you actually need By using DefaultDynamics instead of AbstractDynamics all the methods are programmed for doing nothing so it is only necessary to define the one you really need not all the methods needs to do something When calling each method this variable is defined and appropriately bound in the context time contains the duration since the previous transition remember that these methods are called in a given cycle and the M DEVS bus advances time from a cycle to another The following methods are defined for accessing the incoming influences e getAllInfluences to get the list of incoming influences in any order e getInfluence String name to get the list of incoming influences with the given name It is used to control the order in which to handle the incoming influences e getInternalInfluence to get the incoming internal influen
14. from the bug to the population in order to be able to inform it of the size changes see figure 7 15 The only change concerns the bug population to create this new link and the behavior of each individual bug It touches e the initialization of the bug population for i in range nbBug p port bug i addPort p StupidMode12 Bug parameters pair no i linkPort portRef p position portRef observer linkPort portRef p population self sendProbe objectLayers bug size sendLogical observer populationSize nbBug where a new port link is created from each bug to the population notice the use of the always defined port self e the initialization of the bug for setting the initial size to 1 sendLogical position randomPlace no size 1 sendLogical population size no size e the Ain of the bug to emphasize the double activity sendInternal 1 moveRandomAndGrow e ert of the bug remains the same CHAPTER 7 SOME EXAMPLES 74 null StupidModel2 BugPosition lt lt Defaultindividualinitializer gt gt seed 5612653141 population space Discrete2DSpaceView out2 observer null StupidModel2 BugPopulation null StupidModel2 Space lt lt Defaultindividualinitializer gt gt lt lt Defaultindividualinitializer gt gt nbLine 20 nbCol 20 Figure 7 16 The concrete model 2 sendExternal pos
15. opening saving and deleting whatever is appropriate depending on the clicked item e on Projects it is possible to manipulate the projects list e on Conceptual models it is possible to manipulate the list of conceptual models creating or opening new models e on Concrete models it is possible to manipulate the list of concrete models creating or opening new models e on Simulation models it is possible to manipulate the list of simulation models Creating one gives access to the list of available concret models in the project Open one allows to load a stand alone simulation model e on a specific conceptual model it is possible to create new categories by entering just their name CHAPTER 2 RUNNING MIMOSA 16 Project browser 0 File Window Help M Projects Graph Categories Influence types w ExampleProject v Conceptual models im o M N Panel SP ET vs e E ConceptualMode 1 A BE ConceptualMode 1 B gt M ConceptualModel2 Y M Concrete models E ConcreteModell v Simulation models E simulation ConcreteModel1 ConceptualModel1 A lt lt NativeState gt gt ConceptualModel1 B lt lt NativeState gt gt Figure 2 6 The project window with a selected conceptual model 2 7 The files Each model is stored in a separate file in XML format with a given extension The available extensions are the following pml the file contains a project as a list of related files aml the fi
16. Existing Farm Person Farm Farmer Farm Plot New Name Herder _ New Existing J Cancel Figure 4 3 The creation dialog for a category the fourth is for creating or selecting categories to draw e the fifth one is the taxonomic relationship the sixth is the semantic relationship finally the seventh is the button to access the push down menu for ma nipulating the grid behavior as already described in 2 4 The first three buttons as well as the last one are always present for each graph editor so it shall not be explained again The figure 4 2 shows the use of a note The Edit menu contains an additional item called Export documentation for generating an HTML file containing the exhaustive description of each de fined category including its associated documentation see 4 2 2 This option is very useful to generate a lexicon of all the notions introduced in the conceptual model 4 2 Category edition 4 2 1 Drawing a category To draw a category in a given place it is enough to click on the corresponding button and then at the desired place or to right click at the desired place to show up the same toolbar as a popu menu new dialog is opened as illustrated in the figure 4 3 This dialog is composed of two parts e the upper part lists all the categories available in all the opened ontologies Selecting one of these and typing either return or pushing the Existing button shall draw the corresp
17. Print prints the content of the current window if applicable it is ap plicable when a graph is displayed Refresh this item is only used if you defined a new meta ontology in a so called plugin and you want to dynamically reload the plugins definitions for further use without relaunching Mimosa Edit this menu provides the contextual editing functionalities provided for the selected window or object This menu does not appear when browsing projects yet Any editor provides at least the following functionalities in addition to the usual cut copy and paste Export documentation to generate an html file describing the con tent of the selected window or item in a human readable form Add to add a new object categories individuals states etc Change to change the name of an object when there is an associated name Edit to edit the structure of an object the structure depends on the object and sometimes includes the associated behavior description Remove to remove the selected object s Clear to remove all the defined objects Window this menu provides quick access to the opened windows One of these is always accessible even if not shown by default Output to display the output window which is a console containing a panel for user specific output a panel where error are displayed and a panel where the traces are displayed 1This possibility is provided to create stand alone models wi
18. a C23 The quantities Llosa ii CL Th mumitss LAS a es a a ee A I List of Figures The welcome window The conceptual model editor as an example of an editor window The category list editor of ontologies The graphical editor buttons The scheduler window The project window with a selected conceptual model The project window with a selected simulation model The project window with a selected category Farmer and plot individuals Farmers owning plots The description of the plot p2 A category hierarchy of plots and people A category hierarchy of plots and people with a relationship The buttons of the ontology editor An annotated category The creation dialog for a category The category graphical form The category editor with the attribute panel The category editor with the inherited attributes The creation dialog for a relation The example of a relation The relations of a category The behavior panel of the category The structure of an M DEVS entity The behavior panel of a simple object
19. anonymous mimosa cvs sourceforge net cvsroot mimosa The latest version is available at the head of the project The tagged versions version Obeta to version1 2 5b can be dowloaded but of course are not fully up to date If you want to be a developer just create an account on SourceForge and send a message to jean pierre muller cirad fr to give the name of your account and to explain what you want to do 2 2 Launching Mimosa Mimosa is written in Java 1 5 and can be run on any platform both hardware and operating system as long as at least Java JRE 1 5 is installed For the time being Mimosa is provided as a folder containing e mimosa jar which is the main program to be launched by typing java jar mimosa jar or by double clicking on it if your OS has Java inte grated in it e a libs folder containing the libraries necessary for running Mimosa e an example folder containing some examples to load within Mimosa for exploring its functionalities e a documentation folder for the documentation it should be soon or later a user s manual this one a programmer s manual and the full javadoc hierarchy CHAPTER 2 RUNNING MIMOSA 10 MIMOSA WELCOME IN MIMOSA Choose your language O fran ais anglais O espagnol Cok Figure 2 1 The welcome window e a plugins folder contains so called plugins which are either hard coded examples or additional dynamical specification paradigms Nowadays all the extensions o
20. gt lt name gt t if the influence has the given name contentOf lt influence gt the list of arguments getAllInfluences the list of incoming influences getInfluence lt name gt the list of influences of the given name getInternalInfluence the internal influence e the various events can be posted with the following functions port lt sym gt n1 Nn creates a port sendExternal lt port gt lt sym gt lt exp1 gt lt exp gt post an external event sendInternal n lt sym gt lt expj gt lt expn gt post an internal event sendLogical lt port gt lt sym gt lt exp1 gt lt exph gt post a logical event reply lt influence gt lt sym gt lt exp1 gt lt expn gt reply to an influence sendProbe lt sym gt lt exp1 gt lt exph gt post a probe A port can be a string or a symbol when there is no indices Finally the structure changes can be made through the following functions portRef lt porti gt lt porth gt creates a port reference pair lt sym gt lt exp gt creates a pair for the parameters parameters lt pair gt lt pair gt creates parameters from the pairs addPort lt portref gt lt category gt lt parameters gt creates a new entity linkPort lt portrefi gt lt portref gt gt links referenced ports removePort lt portref gt removes a references port Appendix B Introduction to Smalltalk Smalltalk is a pure object oriented language but here a
21. have to focus on specifying each of the function of the corresponding M DEVS model The figure 7 6 shows how to define the initialization of the rolling ball In the shown panel the Scripting behavior has been selected which allows to specify the behavior with script languages In this case the Java scripting language has been selected Javalnterpreter Note that we distinguish the attributes and the state of the model The at tributes define the structure of the ball for an external observer and corresponds semantically to the specification of its initial state The state itself changes con tinuously spontaneously or in response to incoming influences In this case the state is created and initialized from the parameters The figure 7 7 shows the code for handling incoming external influences The principle is to loop through the set of influences put in the variable externalInfluences to check its type for each one and compute the state change accordingly Note that after the state change a probe value is issued to update all the possible visualization windows CHAPTER 7 SOME EXAMPLES 65 Nom Abstrait O Super type example RollingBall Attributs Relations Comportement Comportements Unit de temps 1 Sondes Influences entrantes Influences sortantes Comportement Comportements Comportement LanguageState dition du code Javainterpreter y Init instance Initialise Inte
22. i in range nbBug p port bug i addPort p StupidModel4 Bug parameters pair no i linkPort portRef p position portRef observer linkPort portRef p population self sendProbe objectEntity bug i p lt HERE sendProbe objectLayers bug size sendLogical observer populationSize nbBug e the space has to do the same thing for its cells for i in range nbLine for j in range nbCol cell port cell i nbCol j addPort cell StupidModel4 Cell1 parameters pair x i pair y j linkPort portRef cell space self sendProbe cellEntity i j cell lt HERE sendProbe cellLayers food sendLogical observer spaceSize nbLine nbCol e each bug has to send probes when its state changes in its initialization size 1 maxConsumptionRate 1 sendLogical position randomPlace no sendLogical population size no size sendProbe size no size lt HERE and in its growth behavior 007 for inf in getAllInfluences if type inf consumed consumed contentOf inf size size consumed sendLogical population size no size sendProbe size no size lt HERE e and the same for the cells as an example in initialize availableFood 0 maxFoodProduction 0 1 rand newRandom 6538672547 sendLogical space food
23. idModel6 BugPopulation null StupidModel6 Space lt lt Defaultindividualinitializer gt gt lt lt Defaultindividualinitializer gt gt maxFoodConsumption 1 0 nbLine 20 nbBug 20 maxFoodProduction 0 1 nbCol 20 Figure 7 28 The concrete model 6 ANO E Output CategoryChartView Chart name 19 size distribution O Bar chart a o ABarchrt3p Legend Name of X axis size O Line chart Line chart 3D i Name of Y axis nb bugs E PRA M Tooltips Figure 7 29 The parameters for a histogram visualizer CHAPTER 7 SOME EXAMPLES 88 losas Figure 7 30 The control panel 6 specification with the new histogram Running the simulation model The figure 7 30 shows the control panel in which the new histogram is shown after 18 steps 7 3 7 Stupid model 7 This version illustrates the possibility to stop the simulation when some condi tions occur In this case the condition is when a bug gets a certain size Defining the conceptual model The conceptual model structure does not change and remains the one of the figure 7 24 The only difference is for the bug stop the simulation when a given condition arises e in the bug growth i e in 04 the condition is tested and the simulation stopped is met for inf in getAllInfluences if type inf consumed consumed contentOf inf size size consumed sendLogical population size no size sendProbe size no size if size gt 3 lt HERE s
24. in 10 12 3 1 Individuals Individuals are the basic ground level components of an ontology The in dividuals in an ontology may include concrete objects such as people animals tables automobiles molecules and planets as well as abstract individuals such as numbers and words Strictly speaking an ontology need not include any in dividuals but one of the general purposes of an ontology is to provide a means of classifying individuals into categories even if those individuals are not explic itly part of the ontology In Mimosa the concrete model editor is provided for defining the individuals out of the defined categories Only the individuals can actually behave and therefore be simulated In figure 3 1 we have three plots pl p2 and p3 and two people John and Paul The name of the individual is optional and indicated before the The name after the semi colon shall be 18 CHAPTER 3 THE ONTOLOGIES 19 John Farm Farme p2 Farm Plot Paul Farm Farmer H p3 Farm Plot Figure 3 1 Farmer and plot individuals 1p1 Farm Plot ownership lohn Farm Farmer ownership 2 Farm Plot f Paul Farm Farmer a ownership 3 Farm Plot Figure 3 2 Farmers owning plots explained in the following It actually is the name of the category the individual belongs to 3 2 Links For the model to be properly called a structure these individuals usually are linked to each other in some meani
25. initializel initialize gt make interval getParameter interval defrule getExternal getExternal gt sendExternal clocked tick defrule getInternal getInternal interval value gt sendInternal value tick It is no longer maintained because Jess requires a licence which is free for academics but costly for others The library is not provided with the distribution for that reason but can be downloaded from 2 Python scripting The implementation uses the Jython library whose documentation can be found on 7 We are using the possibility in this version of Python to call Java objects with the standard Python syntax Accordingly the variable self is defined as CHAPTER 5 THE BEHAVIOR 45 well as all the variables as in Java and the corresponding methods can be called directly So there is not much difference with Java However for facility we have defined the corresponding functions which can be called directly without using the reference to self Prolog scripting Prolog ia a rule base language with a backward chaining semantics The behav ior is described as a single set of rules of the form lt conclusion gt lt conditions gt The program is run by asking for a conclusion and the program tries to find the possible proofs As in Jess each M DEVS function introduces the time the influences and the function name in the fact based and the rules are fired accordingly until no rule
26. is applicable The run predicate must be defined The example of the clock looks like this run initialize X is getParameter interval asserta interval X run getExternal sendExternal clocked tick run getInternal interval X sendInternal X tick Implemented but not yet fully tested The implementation uses the tuProlog library whose documentation can be found on 4 Smalltalk scripting The implementation uses Athena see the we site 1 which is a lightweight im plementation of Smalltalk for embedded applications and has been fully tested A class called Self has been implemented with a number of class methods to access Mimosa from Smalltalk The list of implemented methods can be found in the appendix B 5 3 3 State charts Coming soon 5 3 4 Further extensions This level being extensible at will by adding further meta ontologies this chapter shall only describe some of them as provided in the first versions of Mimosa How to define new meta ontologies is described in the programmer s manual In this chapter we shall introduce the meta ontologies for object space cellular automata and multi agent systems The objects Most categories have very simple behavior corresponding roughly to what is available in objet oriented programming For the categories it is not necessary to provide the full M DEVS functionality although object orientedness can be CHAPTER 5 THE BEHAVIOR 46 mapped in a
27. occupied by a drop down menu to select the conceptual model from which one wants to instan tiate the individuals and links A concrete model can be drawn from several conceptual models combining various sources of knowledge 48 CHAPTER 6 THE CONCRETE MODEL EDITOR 49 MONS NO pal Figure 6 1 The buttons of the model editor the second one is used to visually draw the control panel for the simulation of the corresponding model Apart from the conceptual model drop down menu the starting point is the tool bar in the upper part of the first graph panel as illustrated in the figure 6 1 where six buttons appear 6 1 the first one is the grabber for selecting an object individual or links in the drawing and is always selected by default the second is the note object to write down documentary comments to associate to individuals the third is the link to associate a comment with an individual the fourth is for creating or selecting individuals to draw the fifth one is the link the sixth is used for creating an output the seventh is a link between an individual and an output to specify where to send the probes finally the sixth is the button to access the push down menu for manipu lating the grid behavior as already described in 2 4 Individual edition 6 1 1 Drawing an individual To draw an individual in a given place it is enough to click on the corresponding button and then at the desired place or to
28. refers to another category it is a relationship and no longer an attribute Therefore an attribute is nothing but a relation to a simple notion like numbers 3 5 Relations An important use of relations is to describe the relationships between individ uals in the ontology In fact a relation can be considered as an attribute whose value is another individual in the ontology or conversely an attribute can be considered as a relationship with another individual a number is also an indi vidual instance of the category of numbers For example in the ontology that contains the Farmer and the Plot the Farmer object might have the following relation ownership Plot This tells us that a Plot can be owned by a Farmer Together the set of relations describes the semantics of the domain In the figure 3 5 a relation has been added accordingly In addition we have also declared that a person can be proprietary of any number of plots One can see that the individuals described in figure 3 2 appear to be instances of the categories described in 3 5 and that their links appear to be instances of the related relations In Mimosa a relation is uni directional and links a category to another with a cardinality The most important type of relation is the subsumption relation is superclass of the converse of is a is subtype of or is subclass of already mentioned in the previous section Another common type of relations is the meronymy relatio
29. right click at the desired place to show up the same toolbar as a popu menu A new dialog is opened as illustrated in the figure 6 2 This dialog is composed of two parts the upper part lists all the individuals available in the selected model Selecting one of these and typing either return or pushing the Existing button shall draw the corresponding individual at the selected place the down part is used to create a new individual with two fields a drop down menu from which to select the category one wants to create an individual from a name field to enter a name which is optional but can be used for documentation purpose CHAPTER 6 THE CONCRETE MODEL EDITOR 50 Existing John Farm Farmer Paul Farm Farmer p1 Farm Plot p2 Farm Plot New Category Crarmiplotent Name optional p3 New Existing Cancel Figure 6 2 The creation dialog for an individual p3 Farm Plot lt lt Defaultindividuallnitializer gt gt surface 15 Figure 6 3 The individual graphical form A rectangle is drawn as illustrated in the figure 6 3 with a name which composed of the optional name of the individual a semi colon and the category name which is itself composed of the ontology name and the category name Under its identification the list of attribute values is available Under the name of the individual there is a stereotype mentioning how the individual can be initialized As it is possib
30. self sendProbe cellEntity i j cell sendProbe cellLayers food sendLogical observer spaceSize nbLine nbCol e each bug and each cell has now this variable defined and therefore it is no longer necessary to initialize it separately Defining the concrete model Given the need to define the maximum food production and consumption for the space and the bug population respectively in addition to the initial bug population the concrete model includes these two new attributes The result is illustrated in the figure 7 25 The control panel is now changed to add three new parameters editor see 7 25 In effect any attribute of the concrete model can be associated to an editor to be put on the control panel if needed CHAPTER 7 SOME EXAMPLES 84 null StupidModel5 BugPosition lt lt Defaultindividualinitializer gt gt null StupidModel5 BugPopulation null StupidModel5 Space lt lt Defaultindividualinitializer gt gt lt lt Defaultindividualinitializer gt gt maxFoodConsumption 1 0 nbLine 20 nbBug 20 maxFoodProduction 0 01 nbCol 20 Figure 7 25 The concrete model 5 UDS M GE Figure 7 26 The control panel 5 with the attribute editors CHAPTER 7 SOME EXAMPLES 85 20 nbBug maxFoodConsumption 0 01 maxFoodProduction M Show grid 1 Spread objects Select cell layer Select object layer Recordimage Figure 7 27 The
31. structures it is what functions or all about The nice thing is that eval lt exp gt transforms the structure produced by the expression into an expres sion and computes its value as well Therefore one can write programs pro ducing programs which are further executed This appendix is not suppose to give a full course on Scheme but just pro vide a summary of the most common definitions for reference including the definitions introduced for use within Mimosa A 1 Control syntax As in any language there are some constructs for the usual control structures the sequence the conditional and the loop define lt symbol gt lt exp gt the definition set lt symbol gt lt exp gt to change the definition begin lt exp1 gt lt expn gt the sequence of expressions if lt exp gt lt expirue gt lt eXPfalse gt the conditional cond lt exp1 gt else the multiple contitional or lt exp1 gt lt expn gt sequence until true and lt exp1 gt lt exph gt sequence until false The loop is more complicated with the form do lt iter gt lt iter gt lt cond gt where iter is a variable of iteration of the form lt var gt lt exPinit gt lt eXPstep gt With a variable name an initialization expression and a step com putation expression the condition expression must be true for stopping the iteration and the corresponding expressions are computed accordingly Finally one must
32. 1458 0327303117579725 1 1958624078576456 0830039785680547 1 375521784511871 097742759877525 1 4189290766508607 1161602974103875 1 4189290766508607 160728367517132 1 4888224390261584 2137016348805285 1 5297326736339627 222928550594191 1 5297326736339627 PRPPRPRPPRPRE PRPRPRPRPRPRP Figure 7 34 The result of the simulating the stupid model 8 Chapter 8 The scheduler This chapter is really about running simulations The concrete models one wants to run are available from the drop down menu on the top of the scheduler window see 8 1 All the models defined in the concrete model editor are shown in this drop down menu to select from Additionally files can be loaded within the scheduler if saved in the scheduler format from the concrete model editor This possibility is offered to deliver turn key models to be run independently of all the previously described editors A concrete model has to be selected from the list on the left The initialize button shall actually generate the simulation model out of the concrete model description The first step shall initialize the simulation model the time shall remain at 0 Further steps shall advance the time depending on the closest scheduled next date In the scheduler menu the first item opens an inspector to visualize the list of all created entities see 8 2 This list is updated during the simulation to reflect the current list of entities Clicking on an entity opens an entity inspe
33. HE BEHAVIOR 44 self sendExternal clocked tick There is one drawback in using Java scripting all the Java types have to be prefixed explicitly by the package name for example java lang Integer instead of simply Integer Conversely one can use undeclared variables pro viding flexibility Scheme scripting The Scheme language is a kind of pure functional language based on lambda calculus The facilities for manipulating symbols and lists make it particularly useful for qualitative and symbolic manipulations much less for numerical com putations We are using the Kawa library 6 fast and complete but with scoping problems as well as JScheme 3 limited and slow but semantically consistent for providing Scheme The documentation for the language itself can be found on the corresponding web site The appendix A provides a short reference to the Scheme language as well as the list of provided functions for calling Mimosa Jess scripting Jess is a rule base language with a forward chaining semantics see 2 The behavior is described as a single set of rules of the form lt conditions gt gt lt actions gt Whenever the conditions are met the corresponding rule is fired and the actions executed In our case each M DEVS function introduces the time the influences and the function name in the fact based and the rules are fired accordingly until no rule is applicable The example of the clock looks like this defrule
34. MIMOSA user s manual Draft version 1 4 0 Jean Pierre M ller CIRAD ES GREEN jean pierre muller cirad fr November 11 2012 l Associated researcher to LIRMM Contents 1 Introduction 6 Tl NewS anae a ia a a cn 7 2 Running Mimosa 9 2 1 Downloading Mimosa 9 2 2 Launching Mimosa 9 2 3 Fhemenus gt is a A a ls at us A be Ee 10 2 4 The editor windows 12 2 5 The scheduler window 14 2 6 The project window 15 241 FPhefileg ros Lam O re de 16 3 The ontologies 18 3h Individuales 224 5 sh HAN EE ARE t e A MALE YS 18 32 Links bou aime a A DORE RAR A ANA a 19 3 32 Attributes e LU Le A A A TA SES ot x 19 34 Cat gories 4 4 4 4 th ai hae aae h p terasse died 4 20 3 07 Relations se a met sie tere docs Boel ada eae rl PR dde 21 4 The conceptual model editor 23 AT Th ditor s sa moek e ee a ed e has Ge nes 23 4 2 Category edition 24 4 2 1 Drawing a category 24 4 2 2 Editing a category 25 4 2 3 Deleting a category 25 4 3 Relation edition 26 4 3 1 Drawing a relation 26 4 3 2 Editing a relation 27 4 3 3 Deleting a relation 28 5 The behavior 29 DL Introduction sac super ia a gee lo h
35. RollingBall SchemeKicker HardRollingBall PythonRollingBall PythonObserver HardObserver PythonClock Figure 2 3 The category list editor of ontologies S L M x S Fy Panneau Figure 2 4 The graphical editor buttons model editor or any of these in the project browser by left clicking on its name By right clicking on the panel one accesses a popup menu where it is possible to add a new model open an existing one change its name or delete it It is highly recommended to create a new model each time one is describing a different structure for modularity and reuse reasons The right panel is editor specific and usually allows multiple views of the same model or parts of it In most cases a graphical view is provided In the figure 2 2 there are three editor panels The shown one is the graphical editor panel The other two are used to edit categories and influence types see 5 2 3 as lists The figure 2 3 shows the list editor where it is also possible to add change the name edit and remove categories On the top of any drawing view there is a toolbar with a number of model specific buttons These buttons are specific and shall be described in the related chapters These editing buttons are also available as a popup menu when right clicking in the drawing area The last button is a drop down menu to manipulate the editor window zooming in and out reducing enlarging or hidding showing the grid for objects align
36. THE BEHAVIOR 42 package example import mimosa state DefaultState public class MyClock extends DefaultState private int interval public int getInterval return interval public void setParameter String name Object value throws EntityException if name equals interval interval Integer value intValue public void dolnstancelnitialize StateSpecification stateSpecification throws EntityException The state is composed of a single variable recording the interval between two ticks from the attribute values And the dynamics is as follows package example import mimosa dynamics AbstractDynamics import mimosa scheduler EntityException import mimosa state State public class MyClockDynamics extends DefaultDynamics public void doGetExternal State self throws EntityException self sendExternal clocked tick public void doGetInternal State state throws EntityException self sendInternal Clock self getInterval tick in which we declare a variable to use the interval between two ticks stored in the state the 7 function which signals an output after the given interval and Aert where a single influence is sent to the port Notice the use of DefaultDynamics for not defining all the methods when only two are necessary Of course it is not enough to write the code This code has to be known by Mimosa In order to do that you have to create an XML file in which Mim
37. aceView out10 seed 561 2653141 A population space 4 null StupidModel8 BugPopulation null StupidModel8 Space lt lt Defaultindividualinitializer gt gt lt lt Defaultindividualinitializer gt gt maxFoodConsumption 1 0 nbLine 20 nbBug 20 maxFoodProduction 0 1 nbCol 20 Figure 7 32 The concrete model 8 id size contentOf inf bugSizes id size sendProbe objectState bug id size size min max bugSizes 0 mean 0 for size in bugSizes mean size if size lt min min size elif size gt max max size mean mean len bugSizes sendProbe sizes min mean max Defining the concrete model It is enough to add to the concrete model a new probe observer for the output of the probe information to a file A probe observer is already defined for this purpose and described in the section 6 3 2 The resulting concrete model is illus trated in the figure 7 32 The probe observer can be parameterized as illustrated in the figure 7 33 Here we choose to not output the dates of occurrence Running the simulation model The result of the simulation on 10 steps is shown in figure 7 34 CHAPTER 7 SOME EXAMPLES 91 AAO Output ProbeOutput sizes File name StupidModel8 out O Save probe name e 1 Save global time Sons 1 Save local time Figure 7 33 The file output parametrization 0043141737958714 1 0431417379587131 0229371913650902 1 12415345046
38. age apart from the structure of the model to simulate as described in the previous stage an important work consists in deciding which attributes can be considered as fixed parameters which ones can be manipulated by the user how to output the observables of the model plots grids databases statistical tools etc the simulation stage it consists in running the simulations themselves by creating the simulation model to run as a set of entities linked through ports by connections by associating the means to specify the input pa rameters and to handle the outputs of the simulations and by actually simulating it lIt is the french acronym for M thodes Informatiques de MOd lisation et Simulation Agents computer science methods for agent based modeling and simulation CHAPTER 1 INTRODUCTION 7 In the following we shall describe these stages in turn with the associated concepts But before we shall shortly introduce how to run the system Mimosa is also implemented to be multi lingual For the time being only english french and spanish are provided The user s manual is only in english Most of the explanations still apply even if the menus and titles are not the same 1 1 News The changes from version 1 3 0 are the following 1 2 3 4 5 6 a fair number of bug fixes the project editor is now easier to use the full debug of the scripting languages including Smalltalk a first introduction of the prolo
39. ain which could be as fine as picoseconds and therefore mapped on natural numbers As already mentioned an M DEVS entity only considers durations In addition these durations can only be expressed as integers When simulating an M DEVS entity a local time is deployed The creation of an M DEVS atomic model either at the start of the simulation or during it defines the origin of the local time 0 All the durations are added up generating a local date as an integer In particular this local time is used to compute the durations transmitted to the M DEVS entity A step further the M DEVS bus defines a global time The origin of the global time 0 is the start of the simulation initialization always occurs at the CHAPTER 5 THE BEHAVIOR 39 global time 0 The M DEVS entity local times are mapped to the global time in two ways e the origin of the local time is situated in the global time at the global time of creation of the M DEVS entity e the ratio between the local time grain and the global time grain is given The global time grain is assumed to be the smallest possible grain able to take into account the grain of any other atomic model as an integral multiplier of the global grain Still at this stage the time is a natural number without dimension without unit The correspondence between this time and the real time where the origin of simulation corresponds to a real date and the global grain has a unit pi cosecond hour
40. aph panel for graphical editing e the list panel for editing the ontology as a list of definitions a kind of dictionary e the list panel of influence types to be explained in the section 5 The list panel is the reference to know all the categories defined in the edited conceptual model In effect a category may not appear in the graph panel Conversely a category may appear several times in the graph panel as well as categories from other conceptual models The rational behind this behavior is that the drawing hence the graph panel must have an explanatory power not only a definitory one and therefore any drawing clarifying the explanation should be possible We shall concentrate on the graph panel which is neverthe less easier to use for defining the categories and the relationships The starting point is the tool bar in the upper part of the window as illus trated in the figure 4 1 where seven buttons appear e the first one is the grabber for selecting an object category or relations in the drawing and is always selected by default e the second is the note object to write down documentary comments to associate to categories e the third is the link to associate a comment with a category SD S E Panneau Figure 4 1 The buttons of the ontology editor 23 CHAPTER 4 THE CONCEPTUAL MODEL EDITOR 24 Farm Person 77 lt lt NativeState gt gt age Integer name String Figure 4 2 An annotated category
41. ce To program each functionality a number of methods are defined by cate gories e to manipulate random generators public Random newRandom public Random newRandom long seed public boolean newBoolean Random rand public int newInt Random rand int max public double newDouble Random rand e to easily create ports and port references public Port port String name int index 2it is necessary to hide which kind of generator is used Currently the Mersenne Twister random generator is known as one of the best and provided in Mimosa CHAPTER 5 THE BEHAVIOR 40 public Port portRef Port port e to manipulate the influence content Object contentOf Influence influence which returns either null if there is no content or an array of objects as defined in 5 2 3 Object list Object objects to create an array of ob jects as a content or sub content Object array Object objects to create an array of ob jects as a content or sub content Object object T i where T is one of the Java simple types short int etc to encapsulate them within the corresponding class instance Short Integer etc T toT Object o where T is one of the Java simple types short int etc to unbox them from the corresponding class instance Short Integer etc e to get the initial value of a parameter public Object getParameter String name e to post an influence at a given po
42. control panel Running the simulation model The figure 7 31 shows the control panel in which the three editors for the pa rameters are displayed and can be changed before initializing the model after initialization the changes in the parameter values are no longer taken into ac count until the next initialization 7 3 6 Stupid model 6 This version still does not add any new behavior but provides the possibility to display an histogram of the bug sizes Defining the conceptual model The conceptual model structure does not change and remains the one of the figure 7 24 The only difference is for the bug population to compute a size distribution and to signal the changes of this distribution to any interested probe observer CHAPTER 7 SOME EXAMPLES 86 e in the bug population two data structures are added one for recording the bug sizes and the histogram itself here with 10 bins import math bugSizes sizeHisto for i in range 10 sizeHisto append 0 sendProbe category i size 0 for i in range nbBug p port bug i bugSizes append 0 addPort p StupidModel6 Bug parameters pair no i pair maxFoodConsumption maxFoodConsumption linkPort portRef p position portRef observer linkPort portRef p population self sendProbe objectEntity bug i p sendProbe objectLayers bug size sendLogical observer pop
43. ctor to monitor what is going on in the given entity see figure 8 3 The panel is divided in four panes e the first pane lists the current parameters of the entity and their values e the second pane is the list of current ports with the list of entities their are associated to the third pane is used for managing the probe observers finally the fourth pane displays the warning messages when necessary Scheduler 0 Add concrete model DefaultModel Scheduler 1 Tracing 1 Checking Graphing State Current date Stop date UNKNOWN 0 0 Reset initialize Step Y Run Stop Y Close Figure 8 1 The scheduler window 92 CHAPTER 8 THE SCHEDULER 93 Inspector 0 Entity list StupidModel1 Space O StupidMode 1 BugPopulation 2 StupidModel1 BugPosition l StupidModel1 Cell 3 StupidModel1 Cell 4 StupidModell Bug 15 StupidModel1 Cellm6 StupidModel1 Bug 7 StupidModel1 Cell 8 StupidModel1 Bug 9 StupidModel1 Cell 10 StupidModell Bug 11 StupidModell Cell 12 StupidModel1 Bug 13 StupidModel1 Celk 14 StupidModell Bug 15 Stupid Modell Cell 16 StupidModell Bug 17 StupidModell Cell 18 StupidModell Bug 19 StupidModel1 Cell 20 StupidModel1 Bug 2 1 StupidModel1 Cell 22 StupidModell Bug 23 StupidModel1 Cellm24 StupidModell Cell 25 StupidModell Cell 26 StupidModel1 Cell 27 StupidModel1 Cell 28 StupidModel1 Cell 29 StupidModell Cell 30 Figure 8 2 The main inspector wind
44. d by its links but also by the values of its attributes interpreted as the specification of the initial state of the simulation By editing an individual the dialog of the figure 7 9 appears where you can change the name of the individual optional trace or untrace the individual define or change the attribute values 2while tracing in the scheduler traces the posted and sent influences tracing an individual traces the call to the M DEVS functions CHAPTER 7 SOME EXAMPLES 66 AAO 1 Nom JavaRollingBall Abstrait O Super type example RollingBall Attributs Relations Comportement Comportements Unit de temps 1 J Sondes Influences entrantes Influences sortantes Comportement Comportements Comportement LanguageState dition du code Javalnterpreter 3 Init instance Initialise Interne Externe Confluent gt for Externalinfluence influence externalinfluences if influence is kick x0 x0 vx time y0 y0 vy time java util List lt Object gt direction contentOf influence vx vx self toDouble direction get 0 vy vy self toDouble direction get 1 self sendProbel state self list xO yO vx vy System out printin RollingBall kicked Figure 7 7 The rolling ball category with the external transition panel AAA Mod le ExampleModel DefaultModel example E s E S pez NX Panneau Lexample HardR
45. d selecting Remove from the Edit menu or by right clicking on it and selecting Remove in the popup menu It is asked whether the individual must be removed from the model e if yes the individual is removed both from the drawing and the list of existing individuals defined in the model e otherwise only the drawing is removed but the individual remains as an existing individual 6 2 Link edition 6 2 1 Drawing a link To draw a link in a given place it is enough to click on the corresponding button and then from an individual called the source individual to another one called the target individual or to right click at the desired place to show up the same toolbar as a popu menu new dialog is opened as illustrated in the figure 6 5 This dialog is composed of the list of available relations between the two selected individuals as defined in the corresponding category of the source in dividual Depending on the arity of the relation i e the number of indices to fully specify the relation as many text fields are displayed underneath to enter the indices values In the figure 6 5 the relation is of arity 1 so only one index must be specified The arrow from the source individual to the target individual is annotated by the relation name as shown in the figure 6 6 The index values are written between parenthesis CHAPTER 6 THE CONCRETE MODEL EDITOR 52 pl Farm Plot ohn Farm Farmer o
46. d simulation the structure is often understood as a composition of models each model computing a function to produce outputs outgoing events or values from inputs incoming events or values Of course this composition reflects the structure of the system one wants to model but no discourse on how to describe a system structure is explicitly given On the other hand Artificial Intelligence has focused part of its theories on how people describe the reality This part of Artificial Intelligence evolved partly under the pressure of the web developments both about its contents and its services into what is called today the description of ontologies The term ontology has its origin in philosophy where it is the name of a fundamental branch of metaphysics concerned with existence According to Tom Gruber at Stanford University the meaning of ontology in the context of computer science however is a description of the concepts and relationships that can exist for an agent or a community of agents He goes on to specify that an ontology is generally written as a set of definitions of formal vocabulary Contemporary ontologies share many structural similarities regardless of the language in which they are expressed Most ontologies describe individuals links categories attributes and relations In this section each of these com ponents is discussed in turn as well as the related Mimosa specification More descriptions can be found
47. e 7 3 The stupid model In 13 1t is proposed as series of simple multi agent models of increasing com plexity to both tach and benchmark the modelling and simulation platforms It is used here to illustrate designing models using Mimosa We shall use the Python scripting language for the details All the examples which follow can be loaded at once by opening the project StupidModel in the example folder This project contains the eight versions 3Sorry if we did not program a panel to visualize trajectories yet CHAPTER 7 SOME EXAMPLES 68 with their corresponding concrete and simulation models Feel free to lokk at and to try each of these 7 3 1 Stupid model 1 A number of bugs are randomly positioned in a toroidal grid and move at fixed time steps into a random position within a 4 cells distance Defining the conceptual model As suggested in 9 we propose for conceptual modeling to extensively use the notions of wholes parts and relations Hence bugs are parts of a bug popu lation cells are part of a space and a position is a relation mapping the bug population within the space Accordingly we propose a conceptual model with five categories BugPopulation is a population of bugs of which attribute is the number of bugs Bug isa bug with a given id and of course a part of a bug population Space is a set of cells of which attributes are the number of lines and columns Cell is a cell with given x y coordina
48. e 5 1 The behavior panel of the category e the initialize panel describes how to initialize the state of the individuals each time the simulation is setup The initial state can be retrieved from a data base or set the same for all the individuals e the dynamics panel is used to specify the state evolves over time Once again it can be a differential equation a state machine or whatever suits the chosen state structure In the figure 5 1 the chosen way to specify the behavior is through a state chart as specified in UML called StateChartState Therefore a corresponding state chart editor is shown To understand the offered possibilities and more importantly the behav ior one can expect from these various formalisms i e state initialization and dynamics specifications it is necessary to go down to the ground and expose a little bit of the underlying machinery This is done in the following section Thereafter we shall introduce some already existing formalisms and their cor responding specifications 5 2 The operational semantics Globally the underlying machinery is nothing but a discrete event simulation system The running model is made of entities sending time stamped events which are delivered to their target entities at the specified dates possibly gener ating new time stamped events and so on The scheduler is in charge of ordering the events by their time stamps and to execute them in order The only thing to specify is
49. e M ller and Pierre Bommel An introduction to UML for mod eling in the human and social sciences volume Agent based Modelling and Simulation in the Social and Human Sciences chapter 12 Bardwell Press 2007 Steve Railsback Steve Lytinen and Volker Grimm Stupidmodel and ex tensions template and teaching tool for agent based modeling platforms 2005 Bernard P Zeigler Tag Gon Kim and Herbert Praehofer Theory of Mod eling and Simulation Academic Press 2000 109
50. e definition of the relation name cardinality and type can be either drawn through the graphical editor or entered in the category editor dialog as in figure 7 3 If the relation are defined by the category editor they will not show up in the graphical editor They can be visualized by drawing an arc and specifying an existing link as shown in figure 7 4 At that stage the structure of the conceptual model i e the ontology is entirely defined the categories attributes and relations 7 2 2 Defining the dynamics For defining the behavior you have to define e the incoming and outgoing influences 1To put the seed as a parameter is recommended if one wants to control the outcome of the simulation i e to produce exactly the same result for each simulation CHAPTER 7 SOME EXAMPLES 63 AAA Nom RollingBall Abstrait E Super type E example EntityType Attributs Relations Comportement D clarations d attributs Attributs locaux Attributs h rit s O Cl Cardinalit Type yo 1 Double vx 1 Double vy 1 Double 1 Double x0 Figure 7 2 The conceptual model for a rolling ball with the attribute panel ARA Nom RollingBall Abstrait B Super type example EntityType Attributs Relations Comportement es D clarations de relations Relations locales Relations h rit es O Cl Cardinalit Type observer a Observer Figure 7 3 The rolling ba
51. e two possible port names lt name gt or lt name gt 0 are equivalent Therefore the index is optional if the cardinality is 1 If the reader perceives some relationship between a port and a link he is right We are here using the vocabulary used in the modeling and simulation community which is unrelated to the vocabulary used in the ontology com munity As for individuals generating M DEVS entities the links are used to produce the initial interconnection topology as ports 5 2 3 The influences An influence is an event which is transmitted between two M DEVS entities In Mimosa we also distinguish between influence types and influences as instances of influence types CHAPTER 5 THE BEHAVIOR 34 The influence types are just names but must be declared These names are unique in a given conceptual model or ontology This type level is not really useful at this stage but provides a provision for further typing like the declaration of the arguments to be used for connectivity with other buses like HLA or CORBA where the type of transmitted information has to be declared The influences are instances of influence types For the time being they have e a name which is the name of the corresponding influence type e a content which is either empty or an array of elements For ensuring communication between entities possibly written in various lan guages and in particular in various scripting languages a standard and limited fo
52. economic space defines the monetary value dimension APPENDIX C THE STRUCTURE PACKAGE 106 The physical dimension space referenced by PhysicalSpace PHYSICAL_SPACE is defined with the seven basic physical dimensions e PhysicalSpace LENGTH with the symbol L e PhysicalSpace MASS with the symbol m e PhysicalSpace DURATION with the symbol t e PhysicalSpace ELECTRIC_INTENSITY with the symbol J e PhysicalSpace TEMPERATURE with the symbol 0 e PhysicalSpace AMOUNT_OF_SUBSTANCE with the symbol J e PhysicalSpace LUMINOUS_INTENSITY with the symbol n The economic dimension space referenced by EconomicSpace ECONOMIC_SPACE defines only one dimension e EconomicSpace MONETARY_VALUE with the symbol V Any dimension space needing one and or the other can be build on them using e g e new DimensionSpace PhysicalSpace PHYSICAL_SPACE EconomicSpace ECONOMIC_SPACE The dimensions A dimension is an instance of the class Dimension with the following methods multiply Dimension dimension creates a new dimension as a product of two dimensions divide Dimension dimension creates a new dimension as a quotient of two dimensions power int i creates a new dimension as a power of a dimension root int i creates a new dimension as a root of a dimension isDimensionless true if the coordinates in the dimension space are all 0 isAtomic true if the dimension is atomic sameSpace Dimension dimension true if the two dimensi
53. ed attributes and on the right the list of attributes which have been chosen as state variables In the bottom part one can see the chosen scripting language the chosen incoming influence and the associated code The arguments of the influence if any are stored in the variable arguments as a list The spaces Coming soon The cellular automata Coming soon The multi agent systems Coming soon 3It is assumed that a state variables always has an initial value to be set from the corre sponding attribute CHAPTER 5 THE BEHAVIOR ANA Nom SimpleObject Abstrait oO Super type DefaultOntology EntityType Auributs Relations Comportement Comportements Unit de temps 1 Sondes Influences entrantes Influences sortantes Comportement Comportements Comportement SimpleObject z z m gt gt b lt GB Javalnterpreter calli if a gt 0 sendExternal link inf list c lo ale Figure 5 3 The behavior panel of a simple object Chapter 6 The concrete model editor At this stage the conceptual model has been completely defined both with its structural part the ontology properly speaking with the categories their attributes and their relations and its dynamical part by specifying in a way or another the dynamics of the individuals specified by each category The concrete model editor shall use these definitions
54. ed on durations Although complicated at the first sight the logics is very simple e Meri and feyt are the functions to issue the events Aest and to handle them deez It corresponds straight away to the intuitive event based mechanism as explained in the introduction The events are produced when 7 elapsed since the last transition e Min and fint are the functions for specifying the spontaneous behavior i e what the box does Aint when 7 and how dine e Nog and djog are used to propagate information Azog and make compu tations based on this information djo e Asi specifies the possible modifications in the interconnection topology see below e Aprobe Specifies the observations to provide at each cycle Mimosa implements a unique so called M DEVS bus which is a set of M DEVS entities with interconnected ports More precisely a M DEVS bus is a pair lt E links gt where E isa set of M DEVS entities links is a mapping from E x O into E specifying a mutable interconnection topology For simulation the M DEVS bus runs in cycles Each cycle corresponds to a certain date where everything happening at that date is propagated through all the M DEVS entities At each cycle CHAPTER 5 THE BEHAVIOR 33 1 each model is asked for its 7 Let min be the smallest value 2 the global time is advanced by min Let e C be the set of models with the same minz e C C be the set of models producing
55. ed to create a new relation with three fields a name field to enter a new name which must be unique within the source category a cardinality field to specify whether the relation can reference one several or any number of objects of the given target category The arrow from the source category to the target category is annotated by all the relevant information as shown in the figure 4 8 Additionally the means that each of these links can be drawn with any number of plots The list of defined relations for a category also appears in the relation panel of the category editor as shown in the figure 4 9 A relation can be added or removed directly from this panel but the added relations shall be drawn only if requested as an existing relations The subsumption or taxonomic relationships is a particular case where noth ing need to be specified but the source and target categories 4 3 2 Editing a relation A relation can be edited by double clicking on it or by selecting it and selecting Edit from the Edit menu or by right clicking on it and selecting Edit in the popup menu The same editor appears as for creating it CHAPTER 4 THE CONCEPTUAL MODEL EDITOR 28 4 3 3 Deleting a relation A relation can be deleted by selecting it and selecting Remove from the Edit menu or by right clicking on it and selecting Remove in the popup menu It is asked whether the relation must be removed from the model e if y
56. ell it is on def put id x y bug port bug id newCell port cell y spaceWidth x if id in positions oldX oldY positions id oldCell port cell oldY spaceWidth oldX removePort portRef population bug cell removePort portRef space oldCell bug positions id x y linkPort portRef population bug cell portRef space newCell linkPort portRef space newCell bug portRef population bug sendProbe dPosition bug id x y It illustrates the full power of the link references to remove and create links dynamically throughout the simulation CHAPTER 7 SOME EXAMPLES 79 Discrete2DSpaceView out3 null StupidModel3 BugPosition lt lt Defaultindividualinitializer gt gt seed 5612653141 lt population space null StupidModel3 BugPopulation null StupidModel3 Space lt lt Defaultindividualinitializer gt gt lt lt Defaultindividualinitializer gt gt nbLine 20 nbCol 20 Figure 7 20 The concrete model 3 Defining the concrete model In order to account for the cell food availability the concrete model is slightly changed by adding a link between the space and the visualizer in order to also monitor the changes in the food availability The result is illustrated in the figure 7 20 The figure 7 21 shows the definition of the color codes parameterizing the visualizer
57. ence gt the list of arguments getAllInfluences the list of incoming influences getInfluences lt name gt the list of influences of the given name getInternalInfluence the internal influence 101 APPENDIX B INTRODUCTION TO SMALLTALK 102 e the various events can be posted with the following functions where the arguments are expected to be an array port lt name gt index lt ind gt sendExternal lt port gt name lt sym gt lt sym gt withArguments sendExternal lt port gt name sendInternal n name lt sym gt lt sym gt withArguments lt port gt name lt sym gt lt port gt name lt sym gt withArguments lt sym gt lt sym gt withArguments sendInternal sendLogical sendLogical reply lt influence gt name reply sendProbe n name lt influence gt name lt sym gt lt sym gt withArguments sendProbe lt args gt A port can be a string or a symbol when there is no indices Finally the structure changes can be made through the following functions portRef lt args gt pair lt sym gt with lt exp gt parameters lt pairs gt addPort lt portref gt type linkPort lt portref gt to removePort lt portref gt die lt category gt parameters lt portref2 gt lt args gt lt args gt lt args gt lt args gt lt parameters gt creates a indexed port post an external event 1 post an external event post an internal event v
58. ened outputs files or data bases after the simulation has been stopped Each scheduler window is associated to its own thread so there is a pos sibility of having several scheduler window opened to run several simulations simultaneously 2 6 The project window It is possible to group a set of related models conceptual concrete and simula tion models into a single project Each model is saved in a separate file and the project records the set of files which are grouped together The project window maintains a list of loaded projects each project being composed of conceptual concrete and simulation models The figure 2 6 illustrates the structure of the project window The left vertical pane contains a hierarchy made of a list of projects each one made of a list of conceptual concrete and simulation models respectively Each conceptual model contains a list of categories Each concrete model contains a list of individuals and outputs By left clicking on each of these items one has access to a corresponding editor if available in the right part of the window It is the case for a conceptual model see figure 2 6 or for a single category see figure 2 8 Left clicking on a simulation model gives access to the panel for running the simulation as well as its associated control panel see 8 The resulting display is illustrated in the figure 2 7 Right clicking on an item in the left pane gives access to a conceptual menu for creating
59. eptual model 1 7 12 The concrete modell 7 13 The concrete model 1 control panel 7 14 The simulation model 1 control panel 7 15 The conceptual model 2 7 16 The concrete model 2 7 17 Defining the visualization of the bug size 7 18 The simulation model 2 control panel 7 19 The stupid conceptual model 3 7 20 The concrete model 3 7 21 Defining the visualization of the food availability 7 22 The simulation model 3 control panel 7 23 Choosing a probe observer for cells and bugs 7 24 The stupid conceptual model 5 7 25 The concrete model5 7 26 The control panel 5 with the attribute editors 1 21 The control panels dot a an ere ee EOE 7 28 The concrete model6 7 29 The parameters for a histogram visualizer 7 30 The control panel 6 specification with the new histogram 7 31 The control panel with the histogram 7 32 The concrete model8 7 33 The file output parametrization 7 34 The result of the simulating the stupid model 8 8 1 The scheduler window 8 2 The main inspector window
60. es the relation is removed both from the drawing and the list of rela tions defined for the source category e otherwise only the drawing is removed but the category remains un changed A relation can also be removed from the relation panel of the source category editor If it is deleted this way all the drawings of the relation shall disappear as well Chapter 5 The behavior 5 1 Introduction For each category one can associate a specification of the behavior of the cor responding individuals Basically it is made by selecting a way to specify the behavior a state machine a markov process a differential equation and the list is extensible at will and then by specifying the behavior according to the selected way the states and transitions for a state machine the states and tran sition matrix for a markov process etc The behavior itself is defined by the specification of the state how the state evolves the state dynamics and how to initialize it The behavior is specified by opening the category editor as shown in figure 5 1 and to select the behavior panel At the top there is the specification of the multiplier between the global time grain and the local time graint Further explanation on the representation of time in Mimosa can be found in section 5 2 5 Just under it there is drop down menu to choose the formalism to use conditioning the possible states initializations and dynamics This list is extensible thr
61. f Mimosa are distributed in this form When launching Mimosa a first window is opened to choose your language see figure 2 1 The window shall appear in your operating system language as well as the choice by default However depending with whom you are working any other available language can be selected Thereafter a number of windows appear see 3 In the top the windows for editing the conceptual respectively the concrete models In the bottom left the window to control the scheduler and in the bottom right an output window where a number of information are displayed The next section describes the menus in detail If the latter is the standard and most common use of Mimosa a number of options are now available for running it and available from the command line e java jar mimosa jar launches Mimosa as described before e java jar mimosa jar xxx sml launches Mimosa with only the sched uler and output windows open and the given simulation model already loaded This option is used for delivering turn key models e java jar mimosa jar nw xxx sml launches Mimosa without the user interface and can be used for running simulations in batch mode One can type java jar mimosa jar h to know about further options 2 3 The menus Four menus are provided File this menu provides access to all the functionalities related to the window which is active or to open new windows New is used to open any of the following new windows
62. f the series is optional when there is only one series to display A time chart view is a probe observer that visualizes series of which values depends on time It is parameterized by CHAPTER 6 THE CONCRETE MODEL EDITOR 56 the chart name e the x axis name e the y axis name e whether it has a legend or not e whether it has tool tips or not e whether time is local or global e whether time is used straight or converted to date It understands probes of the form time lt seriesName gt lt value gt where the name can be any string and the value is expected to be a double or convert ible to a double The name of the series is optional when there is only one series to display Discrete2DSpaceView A discrete 2D space view is a probe observer that visualizes spatial information The space is considered continuous by default it becomes discrete when the number of lines and columns hence of cells is specified The cells are considered to have layers corresponding to attributes Only one cell layer can be displayed at a time The space is populated by objects of various types Each object type also has layers Each object is referenced by an id which is unique for each type The parametrization is as follows e one can specify for any cell layer value range a color e one can specify for each object type layer value range a shape a color and a shadow color It understands probes of the forms e csize lt widt
63. for a license python smalltalk and prolog not fully tested yet e with a state transition diagram where the conditions and actions can be specified in one of the scripting languages mentioned before e with any higher level mean of specification as markov processes etc de pending on the availability of the corresponding plug in These various technics shall be described in turn in the next sections 5 3 1 Programmatic specification This section is more appropriate for the programmer s manual but is included here to introduce the basics which are made available in the other ways of specifying the model behavior With your favorite Java IDE for example Eclipse http www eclipse org create a new project with a package let s call it example in which you have to create a number of classes e asubclass ofmimosa state DefaultState describing your behaviour state In case of a parameterizable formalism the resulting class shall be con sidered as the name of your formalism e for each particular way you want to provide for initializing your state define a subclass of mimosa init AbstractStatelnitializer e for each particular way you want to provide the dynamics of your state how the state changes define a subclass of mimosa dynamics AbstractDynamics or mimosa dynamics DefaultDynamics For example for defining a new behavior called MyState the result is a file with the following content CHAPTER 5 THE BEHAVIOR 37
64. for further manipulation 6 3 2 Editing an output An output can be edited by double clicking on it or by selecting it and selecting Edit from the Edit menu or by right clicking on it and selecting Edit in the popup menu The output editor dialog 6 10 shows up with two parts e a drop down menu to choose the kind of output e a panel which depends entirely on the kind of output In the figure 6 10 it is an editor to attribute colors to various probe values for visualization If the output is directed to a file the file should be defined etc The available outputs depend on the behavior associated to the correspond ing individual and are therefore described with the possible dynamical specifi cations However a number of general purpose outputs are provided and shall be described in the following General A number of general probe observers are defined e the probe view to monitor any probes e the probe output for saving the probes to a file A probe view is a probe observer that displays sequentially all the received probes It is useful to monitor any entity A probe output is a probe observer that saves sequentially in a file the selected probes It is parameterized by e the file name e the separator between the saved fields e whether the probe name global time and local time has to be saved too CHAPTER 6 THE CONCRETE MODEL EDITOR 55 e the list of saved probes It can receive any probe of which the arg
65. for providing the user with the possibility to describe as many concrete models as he wants as a set of individuals attribute values and links These individuals attribute values and links are nothing but the instances of the corresponding categories attribute descriptions and relations Their edition shall be described in the sections 6 1 and 6 2 In addition the user must specify what to do with the probes see 5 2 4 Asa reminder the probes are specified in the dynamical description of the categories and must be sent to signal a state change of interest using sendProbe The concrete model editor provides the mean to specify the outputs where one wants to send these probes These outputs can be visual as graphs plots grids etc or can be files databases or even channels to various tools running in parallel like R Excel etc This part shall be described in the section 6 3 Finally the user can visually specify a control panel to be used during the simulation which includes e the visual outputs e the widgets to parameterize the model This latter part shall be described in section 6 4 The concrete model editor is made of two panels e on the left pane there is a list of existing models These models can be created or removed by double clicking in this pane e on the right pane there are two graph panels the first one is a graph panel very similar to the one used for creating conceptual models The top of the panel is
66. g scripting language a new option eclipse which allows the developpement of plugins within eclipse in separate projects the spanish language has been added but still to be improved The changes from version 1 2 3 are the following La 2 TO gt o 6 various options for running Mimosa see 2 2 a new button in the scheduler window see 2 5 a new menu item for generating HTML documentation see 4 1 a new function to stop the simulation programmatically see 5 3 1 a number of probe observers are now described but not all of them see 6 3 2 the stupid model example 13 is in the course of being described see 7 3 The changes from version 1 2 4 are the following 1 2 the smalltalk scripting language is now working see 5 3 2 for space and time efficiency the influences and probes content has been coded as arrays instead of vectors The sections 5 2 3 and 5 3 1 have been slightly changed accordingly the stupid model examples 4 to 8 13 have been added see 7 3 To be read carefully to learn about using Mimosa the implementation packages have been extensively re organized for a clearer understanding of the code This change does not concern this user s manual but only the generated javadoc for the programmers The changes from version 1 2 5 are the following CHAPTER 1 INTRODUCTION 8 1 most importantly the initialization process has been factored out from the dynam
67. ger lt exp gt tests if integer exact lt exp gt tests if exact inexact lt exp gt tests if inexact zero lt exp gt tests if zero positive lt exp gt tests if positive negative lt exp gt tests if negative odd lt exp gt tests if odd even lt exp gt tests if even z1 equality lt a monotonically increasing gt z1 monotonically decreasing lt z1 monotonically non decreasing gt z1 monotonically non increasing abs x the absolute value of the number min z1 the min of the numbers max x the max of the numbers 21 the sum of the numbers 21 the difference of the numbers 21 the product of the numbers z1 the quotient of the numbers quotient n n2 the quotient of the numbers remainder n n the remainder of the numbers modulo n n the modulo of the numbers gcd n the greatest common divisor of the numbers lem n the lowest common multiple of the numbers numerator q the numerator of the rational denominator q the denominator of the rational floor x the floor of the real ceiling x the ceiling of the real truncate x the truncate of the real round x the round of the real real part z the real part of the complex imag part 2 the imaginary part of the complex As well as most transcendant functions A 4 Dotted pairs and lists The most common data structure in Scheme is the dotted pair written
68. h gt lt height gt to specify the extent of the continuous space e dsize lt line gt lt column gt to specify the number of lines and columns of the discrete space e cellLayers lt layerName gt to specify the names of the cell layers e objectLayers lt type gt lt layerName gt to specify the names of the object layers for a given type e cellState lt x gt lt y gt lt layerName gt lt value gt to specify the value of a layer of a cell e objetState lt type gt lt id gt lt layerName gt lt value gt to specify the value of a layer of an object e cPosition lt type gt lt id gt lt x gt lt y gt to specify the position of an ob ject of a given type in continuous space e dPosition lt type gt lt id gt lt x gt lt y gt to specify the position of an ob ject of a given type in discrete space CHAPTER 6 THE CONCRETE MODEL EDITOR 97 Existing Output to Discrete2 DSpaceView out2 New 4 New Existing Cancel Figure 6 11 The creation dialog for an output edge e cellEntity lt x gt lt y gt lt port gt to specify the entity associated to a cell used to associate an observer to it e objectEntity lt type gt lt id gt lt port gt to specify the entity associated to an object used to associate an observer to it The names can be any string the values are expected to be doubles or convertible to a doubles All other
69. he drawing and the list of output edges defined for the model e otherwise only the drawing is removed but the output edge remains un changed 6 4 Control panel definition The control panel editor is used to position the various control panel elements on the control panel The toolbar is shown in the figure 6 12 where apart from the usual buttons we have two main buttons e the green button is used to add an output view to the control panel e the yellow button is used to add a parameter editor to the control panel The figure 6 13 shows a control panel with two parameter editors yellow and one output view green 6 4 1 Drawing an output view To draw an output view in a given place it is enough to click on the corresponding button and then at the place where to put the output view or to right click at the desired place to show up the same toolbar as a popu menu A new dialog is opened as illustrated in the figure 6 14 This dialog is composed of two parts e the upper part lists all the output views available in the selected control panel Selecting one of these and typing either return or pushing the Existing button shall draw the corresponding output view at the selected place e the down part is used to create a new output view with a drop down menu from which to select one of the output view defined in the concrete model graph panel see 6 3 6 4 2 Deleting an output view An output view can be deleted by selecting
70. how each entity behaves i e generates new time stamped events and reacts to incoming events It is the purpose of the next section In the following CHAPTER 5 THE BEHAVIOR 31 Figure 5 2 The structure of an M DEVS entity the events are called influences for obscure another name for historical reasons 8 5 2 1 The model The underlying simulation semantics is based on an extension of DEVS see 14 called M DEVS as a shorthand for Mimosa DEVS Therefore one must understand how M DEVS works in order to master the behavior of the models although most details are assumed to be hidden by higher level of abstractions as suggested in the introduction of this chapter A M DEVS entity is a tuple lt X Y P S O init ext dint Slog dcon Aert Aint Alog Astr probe gt where X is a set of incoming influences which are defined in the incoming infuences panel Y is a set of outgoing influences which are defined in the outgoing infuences panel P is a set of probes which are defined in the probe panel S is a set of possible states as defined in the state panel O is a set of output ports the elements of Y are sent to At this stage we have the structure of an entity regardless of its possible dy namics This entity is represented in the figure 5 2 where the ports are repre sented by black triangles and the arrows shows the incoming influences from the left the outgoing influences through the ports on
71. ics It is now possible to specify how to initialize the simu lation model independetnly of the dynamics For the time being it can be specified at the category level and at the individual level A further improvement concerns the possibility to also vary how to generate the simulation model beforehand 2 the possibility to group a number of models conceptual concrete and simulation into projects as well as browsing these projects in an easier way 3 the mereology browser never completely implemented has been removed definitely 4 the implementation packages have been again extensively re organized for a clearer understanding of the code This change does not concern this user s manual but only the generated javadoc for the programmers 5 the so called plugins are now distributed separatly 6 installers for Windows MacOSX and linux are now provided Chapter 2 Running Mimosa 2 1 Downloading Mimosa Mimosa is a free software under LGPL license and CIRAD copyright The source and code is available on SourceForge If you are only interested in the program itself you can go to the Mimosa site on SourceForge http sourceforge net projects mimosa You just have to follow the link dowload to go to the page where you can download the software How to run it is explained in the next section If you are interested by the software itself or even want to contribute feel free to access it via the CVS server at pserver
72. idModel1 BugPosition 1 population StupidModel1 BugPopulation 2 space StupidModel1 Space 10 position 1 1 position 1 1 position 1 1 position 1 1 i position 1 1 position 1 1 position 1 1 position 1 1 HNEUNNDENONUONNUMENC position 1 1 position 1 1 position 1 1 position 1 1 position 1 1 position 1 1 position 1 1 position 1 1 position 1 1 position 1 1 i position 1 1 position 1 1 PIZZA NINO a Y Clear Message Figure 8 3 The entity inspector window CHAPTER 8 THE SCHEDULER ANA Graph structure a PICKING EES gt Switch to SpringLayout Figure 8 4 The graph of the simulated model 95 Appendix A Introduction to Scheme Scheme is a functional language close to Lisp but with a purer semantics Roughly speaking only two constructs are provided in scheme e the function called procedure in the Scheme community written lambda lt parameters gt lt body gt where parameters is a list of parameter names and body is a sequence of expressions e the application written lt function gt lt arg gt lt arg gt where function is a function as defined before and arg are expressions Of course an expression is either a function or an application This seems overly simplistic but it has been shown that it is enough to e
73. ining the conceptual model The conceptual model will be composed of two categories RollingBall and Kicker The RollingBall is characterized by four attributes two for the initial position x0 and y0 corresponding to xo and yo and two for the speed vx and vy corresponding to vy and vy The Kicker is characterized by one attribute the seed of its random generator used for the time of kicking the ball and the generation of the random vector If we want to visualize the position of the ball the event based nature of the simulation will only be able to provide state changes when the ball is kicked To see the ball rolling between two successive kicks we have to sample the trajectory In order to do that a third category is added to the model to sample the trajectory by asking at each fixed time step to the ball its position The resulting ontology in shown in figure 7 1 In addition you have the definition of three relations e kicked which a relation of Kicker to send a kick to a RollingBall Note that a Kicker can kick simultaneously any number of balls e observer which is a port of RollingBall to send its position to an ob server and it can have as many observers as it wants e observed which is a port of Observer to send a request for position it will always be a logical influence of course The parameters can be edited added changed or removed through the category editor as shown in the figure 7 2 The relations i e th
74. introduce the binding construct to create local variables for various purposes let lt sym gt lt exp1 gt lt exp gt parallel binding let lt symi gt lt exp1 gt lt exp gt sequential binding letrec lt sym gt lt exp1 gt lt exp gt complete binding The main difference is that the association of values to symbols are avalaible from the body alone in the first case directly after the definition and then for the next definitions in the second case and from the start in the third allowing self reference A 2 Booleans There are two booleans t and f which are two symbols which evaluates to themselves Apart from and and or we also have the following functions boolean lt exp gt tests if boolean not lt exp gt the negation eq lt exp gt lt exp2 gt strict equality eqv lt expi gt lt exp2 gt slight extension of strict equality equal lt exp1 gt lt exp2 gt recursive or structural equality APPENDIX A INTRODUCTION TO SCHEME 98 A 3 Numbers Scheme recognizes the integers e g 51236457 rationals e g 6235645 23672573 reals e g 4 6565e 3 and complex numbers e g 3 5i The main distinction is between exact and inexact representations of these The predefined functions are number lt exp gt tests if number complex lt exp gt tests if complex real lt exp gt tests if real rational lt exp gt tests if rational inte
75. ional semantics as described before However as mentioned earlier higher level specifications can be made as various kind of state machines petri nets directly specified differential equations with various means of integration as long as there execution can be mapped in the previously described functions These extensions can be added at will to the system in a way which is described in the programmer s manual When editing a category behavior a number of panes are dedicated for specifying the behavior see the figure 5 1 e the probes to declare which information is dynamically provided during entity simulation CHAPTER 5 THE BEHAVIOR 36 the incoming influences to declare the list of incoming influences the outgoing influences to declare the list of outgoing influences the state pane for specifying the state structure the initialization pane for specifying the initialization method e the dynamics pane to describe the behavior itself The chosen formalism as well as the time grain is specified above these panels Actually the available means for specifying the behavior are as follows e by writing a piece of Java program and declare it to the Mimosa system to make it available in the user interface the so called hard coded or programmatic way e by specifying the behavior of each of the mentioned function using a script ing language Several scripting languages are available java scheme jess unavailable due to a need
76. it and selecting Remove from the Edit menu or by right clicking on it and selecting Remove in the popup menu It is asked whether the output view must be removed from the control panel e if yes the output view is removed both from the drawing and the list of output views defined for the control panel e otherwise only the drawing is removed but the output view remains un changed CHAPTER 6 THE CONCRETE MODEL EDITOR 59 OOO Concrete model StupidModel3 StupidModei3model xml Concrete model Control panel DefaultModel StupidModel3 le A Em Sy Ma E Panes Figure 6 13 The control panel view QOO Existant Discrete2DSpaceView out1 eee Sortie Discrete2DSpac Nouveau Existant Annuler Figure 6 14 The creation dialog for an output view CHAPTER 6 THE CONCRETE MODEL EDITOR 60 Existant StupidModel3 BugPosition indO seed StupidModel3 BugPopulation ind1 nbBug Nouveau Individu Space 6814623 Param tre nbCol Nouveau Existant Annuler Figure 6 15 The creation dialog for a parameter editor 6 4 3 Drawing a parameter editor To draw a parameter editor in a given place it is enough to click on the corre sponding button and then at the place where to put the parameter editor or to right click at the desired place to show up the same toolbar as a popu menu new dialog is opened as illus
77. ition moveRandom no 4 e but the d of the bug is now doing something if type getInternalInfluence moveRandomAndGrow size size l sendLogical population size no size Provision is made for the possibility to have a variety of internal behaviors hence the test e Given the logical influences sent to the bug population these must be handled appropriately by defining the 0 7 of the bug population for inf in getAllInfluences if type inf size id size contentOf inf sendProbe objectState bug id size size Both in initialization and in the dio of the bug population probes are sent for the visualization of the bug sizes Defining the concrete model In order to account for the bug growth the concrete model is slightly changed by adding a link between the bug population and the visualizer in order to monitor the changes in the bug size The result is illustrated in the figure 7 16 The figure 7 17 shows the definition of the color codes parameterizing the visualizer Running the simulation model The figure 7 18 shows the resulting visualization while running the simulation A colored shadow appears and is modified while the bugs are growing It il lustrates the possibility to get the probes from various sources to compose a CHAPTER 7 SOME EXAMPLES 75 Output Discrete2DSpaceView Cell Object Type z Layer
78. le contains a conceptual model made of categories relations as well as initialization and dynamics specifications eml the file contains a concrete model made of individuals links a well as initialization and output specifications sml the file contains a complete simulation model containing alos the initial ization and output specifications to be run stand alone CHAPTER 2 RUNNING MIMOSA nao Project browser 0 File Window Help B Projects w ExampleProject v Conceptual models v ConceptualModell 1 ConceptualModel1 A ConceptualModel1 8 gt ConceptualModel2 v Concrete models A ConcreteModel1 v Simulation models A simulation ConcreteModel1 Scheduler C Tracing Checking State Current date Stop date UNKNOWN 0 0 Reset Initialize Step Run Stop Close Figure 2 7 The project window with a selected simulation model ono Project browser 0 File Window Help ln Projects Abstract w M ExampleProject v M Conceptual models w ConceptualModel1 Super category f 2 ConceptualModel1 B Documentation Attributes gt ConceptualModel2 Documentation Y Concrete models A ConcreteModel1 v Simulation models E simulation ConcreteModel1 Abstract a Relations Dynamics 1 Figure 2 8 The project window with a selected category Chapter 3 The ontologies In modeling an
79. le to describe how to initialize a state for all the individuals in the category behavior part it is possible to specify for each single individual how to initialize it It is usefull when the individual is itself a complicated structure for example when it generates additional individuals before running the simulation 6 1 2 Editing an individual An individual can be edited by double clicking on it or by selecting it and selecting Edit from the Edit menu or by right clicking on it and selecting Edit in the popup menu The individual editor dialog 6 4 shows up with the following parts e the name of the category which cannot be changed e the name of the individual which can be changed at will a panel where one can specify the attribute values a panel where it is possible to select among several ways to initialize the individual and to edit the associated parameters if necessary In this case it has been chosen to initialize it interactively A list of the parameters to be specified is displayed underneath CHAPTER 6 THE CONCRETE MODEL EDITOR ol p3 Attribute Value surface 15 Parametersinitializer C surface Ok Cancel Figure 6 4 The individual editor with the attribute and the initialization panels New Relation type ownership 1 0 New s Existing Cancel Figure 6 5 The creation dialog for a link 6 1 3 Deleting an individual An individual can be deleted by selecting it an
80. ll category with the relations panel ANA Existing kicked New Relation Arit 2 Cardinalit New Existing Cancel Figure 7 4 Definition of an arc from an existing relation definition CHAPTER 7 SOME EXAMPLES 64 Nom Abstrait E Super type example RollingBall Attributs Relations Comportement Comportements Unit de temps 1 Sondes Influences entrantes Influences sortantes Comportement D clarations de sondes Sondes locales Sondes h rit es O Cl Cardinalit Type state 4 Double Ok Cancel Figure 7 5 The rolling ball category with the probes panel e the probes e the M DEVS functions We assume that RollingBall receives kicks and observation requests and sends positions Kicker sends kicks and the Observer sends observation requests and receives positions The checking of the consistency between what is sent or received is currently very loose but can be reinforced by selecting the verify check bon in the scheduler In a future release the possibility to check for model consistency when defining the conceptual model will be enforced at least optionally We shall define two identical probes one for the RollingBall to signal the state change new x0 y0 vx and vy see figure 7 5 and one for the Observer for the ball position each time it receives the actual coordinates These declarative parts of the dynamics being made we
81. ment The figure 2 4 shows the buttons for editing rather complex conceptual models just for illustration Any created object can be edited by double clicking on it On right clicking CHAPTER 2 RUNNING MIMOSA 14 Scheduler 0 Add concrete model DefaultModel 4 Scheduler 1 Tracing C Checking Graphing State Current date Stop date UNKNOWN 0 0 Reset initialize Step Run C Stop gt Close Figure 2 5 The scheduler window on an object one can access a popup menu for editing same as double clicking or deleting the object 2 5 The scheduler window On the top of the scheduler window see 2 5 the list of existing models is provided for inclusion within the list of available models to the scheduler It is also possible to add additional models to simulate by loading them from scheduler specific files This possibility is used when delivering turn key models The bottom of the scheduler window is divided in two vertical panels In the left panel there is the list of existing models as added from the model editor or from files Exactly one model must be selected to be run The right panel is divided in three horizontal panels 1 the top panel has two check boxes for debugging Trace to turn tracing on and off If the trace is on the influences posted and sent are displayed in the trace window Verify to turn verifying on and off If the verify is on all the declara tions
82. n written as part of that represents how objects combine together to form composite objects For example if we extended our example ontology to include objects like Steering Wheel we would say that Steering Wheel is part of Ford Explorer since a steering wheel is one of the components of a Ford Explorer If we introduce part of relationships to our ontology we find that this simple and elegant tree CHAPTER 3 THE ONTOLOGIES 22 structure quickly becomes complex and significantly more difficult to interpret manually It is not difficult to understand why an entity that is described as part of another entity might also be part of a third entity Consequently individuals may have more than one parent The structure that emerges is known as a Directed Acyclic Graph DAG The part of the ontology consisting of the categories attribute descriptions and relations either taxonomic or semantical shall be called the conceptual model The part of the ontology consisting of the individuals their attribute values and their links shall be called the concrete model In the following the editor to create the conceptual model shall be described In addition we shall describe how to specify the dynamics associated to each category Thereafter we shall introduce the concrete model editor Chapter 4 The conceptual model editor 4 1 The editor The conceptual model editor is made of three panels for editing the conceptual model e the gr
83. n of the units implemented in the package structure unit There are several kind of values among which e the atomic values a booleans either true or false a names any string of characters a quantities a value along a dimension a units is a value used as a reference for quantities e the composed values the attributes are sets of heterogeneous values indexed by strings of characters the tuples are sets of heterogeneous values indexed by positive inte gers including 0 APPENDIX C THE STRUCTURE PACKAGE 108 the maps are sets of heterogeneous values indexed by coordinates the sets are sets of homogeneous values with no indexation We will describe the values in turn C 2 1 The boolean The boolean is made of two possible values of class Boolean one is called true the other is called false They are distinguished from the value true and false of a programming language although they can be converted into them for computational purposes C 2 2 The names The names are strings of characters of class Name They are distinguished from the strings String in Java of a programming language although they can be converted into them for computational purposes They corresponds more to symbols in Lisp without the guarantee of unicity C 2 3 The quantities The quantities are values along a dimension of class Quantity They can be converted into an integer or a float or any numerical value
84. n to all molecules Number the category of all numbers Vehicle the category of all vehicles Car the category of all cars Individual representing the category of all individuals Importantly a category can subsume or be subsumed by other categories For example Vehicle subsumes Car since necessarily anything that is a mem ber of the latter category is a member of the former The subsumption relation is used to create a hierarchy or taxonomy of categories down to very specific categories like MaizeFarmer at the bottom Figure 3 4 shows such a hierarchy of categories Usually what is common to a collection of individuals is that they share the same attributes In the figure 3 4 all the people have a name and an CHAPTER 3 THE ONTOLOGIES 21 Farm Person lt lt NativeState gt gt age name Farm Plot lt lt LanguageState gt gt surface cover Farm Herder PETE ownership lt lt LanguageState gt gt lt lt tanguagestate gt gt f cashFlow Figure 3 5 A category hierarchy of plots and people with a relationship age We also assume that each farmer has a cashflow but not a herder By subsumption any farmer and any herder has also a name and an age because they are particular case of Person In Mimosa an attribute has a name a type which can be only a single type short integer long float double string and color and a cardinality to have list of values If an attribute
85. names types and cardinality are checked during simulation It slows down the simulation quite a bit but it is very useful for checking whether the behavior is consistent with the declarations as well as a button to visualize the simulated structure as a graph Cur rently the graph is not updated while running the simulation Therefore the button has to be pushed each time one wants to visualize the current state to be improved later on 2 the middle panel displays the state of the simulation unknown initialized running or stopped and the current date in global time An end date can be entered to specify when to stop the simulation The core simulation system being event based this is NOT a number of steps but really an end date 3 the bottom panel has buttons for controlling the simulation Reset for creating the simulation model and control panel out of its description either file or concrete model Initialize for putting the model in its initial state The current date is always set to 0 CHAPTER 2 RUNNING MIMOSA 15 Run to run the simulation until the provided end date is reached If the end date is less or equal to the current date nothing happens Step to run one cycle of the simulation All the influences scheduled at the next date are executed Stop to stop the simulation before the end date is reached The current cycle is always completed and cannot be interrupted Close to close the possibly op
86. ndom spaceHeight put id x y Notice the possibility to decide the order in which to handle the influ ences Here the size declarations are handled before positioning the bugs CHAPTER 7 SOME EXAMPLES 71 Discrete2DSpaceView out 1 null StupidModel1 BugPosition lt lt Defaultindividualinitializer gt gt seed 5612653141 population space observer null StupidModel1 BugPopulation lt lt Defaultindividuallnitializer gt gt null StupidModel1 Space lt lt Defaultindividualinitializer gt gt nbLine 20 nbCol 20 Figure 7 12 The concrete model 1 although all these influences are received at once For simplicity we do not test whether a cell is already occupied Defining the concrete model Given that the bug population and the space automatically generate the bugs and the cells It is enough to define the bug population the space and the position as well as defining the initial values The resulting concrete model is illustrated in the figure 7 12 In addition we define a probe observer to visualize the position of the bugs in the grid space The Discrete2DSpaceView see 6 3 2 is used for this purpose Finally a control panel containing the Discrete2DSpaceView is defined see 7 13 Running the simulation model Finally the concrete model is used to generate the simulation model for the scheduler The scheduler reset button is used to actually generate the ini tial simulati
87. ng a parameter editor 7 Some examples 7 1 Introductions 2 2 Le Fas ed a tato qe iaa ent 7 2 The rolling ball example 7 2 1 Defining the conceptual model 7 2 2 Defining the dynamics 7 2 3 Defining the concrete model 7 3 The stupid model Jl Stupid modell s ansa ae lb ee eee Et ue 4 7 3 2 Stupid model 2 22 244248 4 ii 7 3 3 Stupid model 3 1 34 Stupid model 4r er ed a c es 7 3 5 Stupid model5 7 3 6 Stupid model 6 7 3 7 Stupid model 7 138 stupid model 8 200 a A el 8 The scheduler A Introduction to Scheme AL Control SyDlax a ra 4 Maa gonea de dus ra e da talas eed A 2 Booleans e ice a Re ee ee ALI Nubes LES E A ta tas dd a A 4 Dotted pairs and lists A 5 Mimosa primitives CONTENTS B Introduction to Smalltalk B 1 Mimosa primitives C The structure package C 1 C 2 The meta data box 4 CAL The typ s x iana e Red aoe a fe ee es ale C 1 2 The dimensions E L 37 The tests 5 dabaa foe ASS SUN A AR A A C 1 4 The values ti ler nn a mue ee The da ta box 4 52 as d RS dr M ee rs ta CA The bool an 44 ds d 5 22 The ames zans a a a oe ae
88. ngfull way In our example the figure 3 2 shows some links between the individuals describing that John is proprietary of pl and p2 while Paul is proprietary of p3 The proprietary link is indicated by the name ownership 3 3 Attributes Individuals in the ontology are described by specifying their attributes Each attribute has at least a name and a value and is used to store information that is specific to the individual it is attached to For example the p2 individual has attributes such as surface 20 cover tree The value of an attribute can be a complex data type in this example the value of the attribute called cover could be a list of values not just a single value CHAPTER 3 THE ONTOLOGIES 20 p2 Trace oO Cl Valeur surface 20 cover tree Ok Cancel Figure 3 3 The description of the plot p2 Farm Person lt lt NativeState gt gt age name Fas en lt lt LanguageState gt gt lt lt LanguageState gt gt cashFlow Figure 3 4 A category hierarchy of plots and people In the figure 3 3 some of the attributes are represented 3 4 Categories Categories are the specification of the common features of groups sets or col lections of individuals They are abstractions over sets of concrete individuals Some examples of categories are Person the category of all people describing what is common to all people Molecule the category of all molecules describing what is commo
89. not be explained further here Finally defining the dynamics would result in the following class package example import mimosa dynamics AbstractDynamics CHAPTER 5 THE BEHAVIOR 38 import mimosa scheduler EntityException import mimosa state State public class MyStateDynamics extends AbstractDynamics public void doConfluentTransition State self throws EntityException public void doExternalTransition State self throws EntityException public void doGetExternal State self throws EntityException public void doGetInternal State self throws EntityException public void doGetLogical State self throws EntityException public void doGetStructural State self throws EntityException public void dolnternalTransition State self throws EntityException public void doLogicalTransition State self throws EntityException AbstractDynamics defines eight 8 methods e public void doExternalTransition State self throws EntityException equivalent to Oct e public void doInternalTransition State self throws EntityException equivalent to fint e public void doLogicalTransition State self throws EntityException equivalent to djog e public void doConfluentTransition State self throws EntityException equivalent to con CHAPTER 5 THE BEHAVIOR 39 e public void doGetExternal State self throws EntityException equivalent to Aert e public void doGetInternal State self throws EntityException
90. oid addPort PortReference name String categoryName Map lt String Object gt parameters creates an entity as an instance of the given category whether it is traced or not and the map of attribute values e void addPort String name String categoryName Map lt String Object gt parameters same as above when there is a simple syntax for the port reference e void linkPort PortReference portRef1 PortReference portRef2 links the port reference to the entities referenced by the second port ref erence creating new links e void linkPort String portRef1 PortReference portRef2 same as above e void linkPort PortReference portRef1 String portRef2 same as above e void linkPort String portRef1 String portRef2 same as above e void removePort PortReference portRef removes the entities from the given port without destroying the referenced entities they kill them selves using die e void removePort PortReference portRef same as above To simplify the specification of the parameters in addPort two additional meth ods are provided e public Pair pair String name Object args for creating a pair parameter name value e public Map lt String Object gt parameters Pair args for creating the adequate map from the pairs For example if we want to program the behavior of a clock which sends a influence named tick to its clocked port at interval time we could have the state defined this way CHAPTER 5
91. ollingBall Figure 7 8 The concrete model as an instance of the conceptual model CHAPTER 7 SOME EXAMPLES 67 Trace z Cl Valeur yo 0 0 vx 1 0 vw 1 0 x0 0 0 Ok Cancel Figure 7 9 The edition dialog for an individual Figure 7 10 The view on the rolling ball state Once all the model has been instantiated and all the parameters defined a further version should also check for the model completeness the user can open the scheduler select the model to run initialize and run it either step by step or in a single run until the end date is reached as described in more details in the chapter 8 In addition a visualization window can be opened For example a pos sible view looks like the figure 7 10 and is updated each time the individuals change The top left panel displays the clock value the top right panel displays KICKED for some time each time the kicker is issuing a kick the bottom left panel displays the rolling ball state updated only when kicked and the bottom right panel displays the actual position of the ball at each time step Such a display cannot be created interactively yet A number of visualiza tion items can be created positioned within a control board and linked to the individuals receiving its probes and using them to update the visualization An editor for such a control panel including the possibility to change the parame ters shall be available in a near futur
92. on model and the control panel The init button initializes the simulation model calling the initialization of each initial entity i e the bug population the space and the position It is at this stage that the bugs and cells are created Finally the simulation model is ready to run see figure 7 14 7 3 2 Stupid model 2 The bugs are allowed not only to move but also to grow starting from a size of 1 increased incrementally by 1 It is visualized as a changing shade color of the bugs in the control panel CHAPTER 7 SOME EXAMPLES 72 Figure 7 13 The concrete model 1 control panel M Show grid 1 Spread objects Figure 7 14 The simulation model 1 control panel CHAPTER 7 SOME EXAMPLES 73 StupidModel2 BugPopulation lt lt Scripting gt gt population 1typidModel2 BugPosition nbBug Integer Ar seed Long 2 1 Space population 1 StupidModel2 Bug position lt lt Scripting gt gt eee StupidModel2 Cell StupidModel2 Space lt lt Scripting gt gt lt lt Scripting gt gt y Integer nbLine Integer x Integer nbCol Integer Figure 7 15 The conceptual model 2 observer Defining the conceptual model The only change in the conceptual model structure is the addition of a relation
93. onding category at the selected place CHAPTER 4 THE CONCEPTUAL MODEL EDITOR 25 Farm Person lt lt NativeState gt gt age Integer name String Figure 4 4 The category graphical form e the down part is used to create a new category with a name field to enter a new name which must be unique within the current ontology A rectangle with either one or two subparts shall be drawn at the selected place 4 4 e the upper part has two lines the first line is the name of the category prefixed by the name of the ontology the second line is the name of the way to define the behavior for this category NativeState is chosen by default and does nothing e the down part is the list of attributes with their specification 4 2 2 Editing a category A category can be edited by double clicking on it or by selecting it and selecting Edit from the Edit menu or by right clicking on it and selecting Edit in the popup menu The category editor dialog 4 5 shows up with the following parts e the name of the category which cannot be changed e an abstract check box to specify whether the category can have instances or not e g most probably in our example there shall not be direct instances of Person but only of Farmer and Herder e the super type i e the category subsuming this category e a panel where one can specify either the documentation the attributes the relations and the behavior see cha
94. ons are defined in the same dimension space As we have seen in the previous section they are created by specific methods of a dimension space APPENDIX C THE STRUCTURE PACKAGE 107 C 1 3 The tests The tests are classifiers to specify sets in intention For the time being the following tests have been implemented Any that specifies the set of all the possible values Comparison that specifies the set of all the values comparable to a given value where the comparison can be the equality inequality less than etc And that conjoins two classifiers that must be simultaneously true Or that conjoins two classifiers that cannot be simultaneously false The two first classifiers are created by using the following static methods of the class Classifier any creates an Any test compare Comparator comparator Value value creates a new Comparison test Two methods are used to conjoin a classifier with another and Classifier classifier creates an And test or Classifier classifier creates an Or test C 1 4 The values The values will be described in the section on the data box However a value is always considered as a classifier specifying the set of values containing only itself Therefore the corresponding test amounts to the equality test C 2 The data box Any data structure is an instance of a value implemented by the class Value They are implemented in the package structure data with the exceptio
95. or magnitudes to describe all the possible physical dimensions the length the duration the mass the electric current the thermodynamic temper ature the amount of substance and the luminous intensity Any other physical dimension can be expressed as a combination of those using multiplication or APPENDIX C THE STRUCTURE PACKAGE 105 power For example the acceleration is a length multiplied by a duration at the power 2 or equivalently a length divided by a duration at the power 2 We generalize the idea by providing the possibility to define an arbitrary set of atomic dimensions like for example the monetary value or the number of apples all the others being combinations of those Additionally let us mention the special case of the dimensionless dimensions obtained when dividing a di mension by itself for example a length by a length as it is the case for ratios and angles for example The dimension spaces Given a set D of atomic dimensions any dimension is a combination of these atomic dimensions using the operations x and the relative integral powers It is easy to see that the operations and behave like the and operations of a vector space of dimension atomic D and the power as the multiplication by an integer Therefore we propose to represent the set of all the possible dimensions as a vector space S K where K Z is the set of relative integers is the vectorial sum and the m
96. or week shall be specified externally It is foreseen to be able to declare this information to the scheduler and use this reference to define in an easier way the time units of the entities It is not yet completely implemented at this stage In summary any M DEVS entity has e a grain the smallest undistinguishable time difference defined implicitly by having durations expressed with integers and explicitly by a multiplier with the global grain e an origin defined implicitly by having the entity life starting at 0 and explicitly by a position of this origin with respect to the global time Although each M DEVS entity has to define the above described functions for specifying its dynamics and therefore to behave properly the user do not necessarily need to know about the existence of these functions For example the user could only provide a differential equation an integration method and an integration step and a proper definition of the 6 and functions shall take care of integrating the equation correctly This is done by specifying once and for all a particular plugin called a formalism The already provided formalisms are described in the following section 5 3 The behavior specification In order to define a formalism for specifying the behavior of an entity the programmer must expect to have to specify each of the mentioned function for proper functioning of the model hence the importance to understand the under lying operat
97. ories defined in the ontology e otherwise only the drawing is removed but the category remains as an existing category 4 3 Relation edition 4 3 1 Drawing a relation To draw a relation in a given place it is enough to click on the corresponding button and then from a category called the source category to another one called the target category or to right click at the desired place to show up the same toolbar as a popu menu A new dialog is opened as illustrated in the figure 4 7 New Relation ownership Cardinality f New Existing Cancel y Figure 4 7 The creation dialog for a relation CHAPTER 4 THE CONCEPTUAL MODEL EDITOR 27 lt lt NativeState gt gt Farm Plot lt lt NativeState gt gt Figure 4 8 The example of a relation ownership Name Abstract Super category Farm Person Documentation Attributes Relations Dynamics Relation declarations gt O Local relations 2 Inherited relations O Relation Cardinality Category ownership Plot ok gt Cancel Figure 4 9 The relations of a category This dialog is also composed of two parts even if in the figure 4 7 only one shows up e the upper part lists all the existing relations available between the two selected categories Selecting one of these and typing either return or pushing the Existing button shall draw the corresponding relation be tween the two categories e the down part is us
98. osa can read the following declarations lt xml version 1 0 gt lt mimosamodule name Example package example gt CHAPTER 5 THE BEHAVIOR 43 lt behaviour notion EntityType implementation MyClock gt lt parameters gt lt parameter name interval cardinality 1 type java lang Integer gt lt parameters gt lt outInfluences gt lt influenceType name tick gt lt outInfluences gt lt behaviour gt lt dynamics state MyClock dynamics MyClockDynamics gt lt mimosamodule gt This XML file contains everything you would have declared through the user interface and additionnaly defines through the package and implementation attributes where to find the corresponding class for the state The dynamics tag declares that MyClockDynamics is one way in this case the only way to define the dynamics of MyClock You then have to create a folder called example to put the jar containing the compiled class to define a file called example config xml containing the content above and to put the whole folder in the plugins subdirectory of Mimosa By trying this example the behavior MyClock will appear in the list of available formalisms In general any new behavior or way of defining behaviors can be added to Mimosa by putting in the plugins directory a folder called xxx with a file called xxx config xml in it with the related XML content and as many jar as necessary Further details as well as the complete syntax of
99. ositioning and moving the bugs around on request The initialization prepares the needed structures and functions populationSize 0 spaceWidth 0 spaceHeight 0 positions random newRandom seed def put id x y positions id x y sendProbe dPosition id x y In particular a function to put a bug in a given place is defined and just adjusts the table of positions and signals the move to any interested observer The only external influence it receives is the request for random moves SO de is defined as for inf in getInfluences randomMove id amount contentOf inf x y positions id incr nextInt random amount if nextBoolean random newX x incr spaceWidth else newX x incr spaceWidth incr nextInt random amount if nextBoolean random newY y incr spaceHeight else newY y incr spaceHeight put id newX newY The move is randomly generated according to the toroidal grid Finally the position must register the population and grid sizes as well as initialize the bugs initial positions by defining dint for inf in getInfluences populationSize populationSize contentOf inf for inf in getInfluences spaceSize spaceHeight spaceWidth contentOf inf sendProbe dSize spaceHeight spaceWidth for inf in getInfluences randomPlace if type inf randomPlace id contentOf inf x nextInt random spaceWidth y nextInt ra
100. ough plugins The behavior panel is itself made of six subpanels e the probe panel is used to define what can be observed from the individu als It is used for displaying what happens during the simulation or saving it to any media for further processing statistics etc The probes shall be described in the section 5 2 4 e the incoming influences panel is used to specify the events the individuals are able to react to They shall be explained in the sections 5 2 1 and 5 2 3 e the outgoing influences panel is used to specify the events the individuals are producing They shall also be explained in the sections 5 2 1 and 5 2 3 e the state panel is used to specify the state structure which can be a vari able a set of possible states a set of attributes a knowledge base or anything depending on how one want to describe the behavior of the in dividuals of this category 1The grain is the smallest difference between any two time measure which can be distin guished 29 CHAPTER 5 THE BEHAVIOR 30 Abstract Abstract Super category DefaultConceptualModel Person Documentation Attributes Relations Behavior State definition Time ratio 1 Formalism StateChartState Probes in influences Out Influences State initialization Dynamics State definition Statechart Logical reaction Os 0 NL Pel class mimosa state language Javainterpreter Ok Cancel Figur
101. outputs 3 Aext is called for each model in C and the outgoing influences are gathered and their destinations are identified using links 4 for each model m in C e if m has simultaneous incoming influences and an internal change dcon is called e if m has only an internal change int is called e if m has only incoming influences dexz is called and all the outgoing logical influences are gathered 5 all the logical and structural influences are dispatched via links by calling Alog Astr and 0109 until there is no logical influences left be careful about possible loops which are not detected 6 all the observations of the state changes are issued caling Aprobe For each individual MIMOSA shall generate a corresponding entity which shall be initialized from the list of its attribute values in a state specific way as specified by init A more formal and detailed account can be found in 11 5 2 2 The ports A port provides a way to connect M DEVS entities together A port can connect an entity to any number of other entities In Mimosa we distinguish between a port and a port name A port name can be a simple name a String designating all the entities linked through the given port or a name with an index with the syntax lt name gt lt int gt designating one of the entities linked through this port In the case the cardinality of the port is one only one entity can be linked through this port Th
102. ow The most important pane certainly is the third pane because it monitors what is going on inside of the inspected entity It is composed of a drop down menu for selecting a probe observer and a panel to display the probe observer when it is displayable By default two probe observers are available e the probe view which displays the probes when received one after the other button to clear the display is available if necessary e the probe output which send the probes to a file When selecting the probe observer a file name as well as a separator string is asked The resulting file can be loaded in excel or any similar tool At each time step it is possible to open a window showing the structure of the simulated model as a graph where each node is an entity and each edge is a connection between the entities The corresponding window is shown in the figure 8 4 and is made of three parts e The upper part is a drop down menu to select the kind of graph manipu lation either transforming for changing the place of the graph zoom it in or out etc or picking for selecting one node and move it on the screen e The graph itself e A button to switch between two algorithms to layout the graph Choose the one which seems more appropriate to visualize the model CHAPTER 8 THE SCHEDULER ADO Entity inspector StupidModel1 BugPosition 1 0 p Parameters seed 5 612 653 141 Ports Key Value self Stup
103. pectively Accordingly e the bug population does initialize each bug for i in range nbBug p port bug i addPort p StupidMode15 Bug parameters pair no i pair maxFoodConsumption maxFoodConsumption lt HERE linkPort portRef p position portRef observer linkPort portRef p population self sendProbe objectEntity bug i p sendProbe objectLayers bug size sendLogical observer populationSize nbBug CHAPTER 7 SOME EXAMPLES 83 ptupidModel5 BugPosition lt lt Scripting gt gt seed Long space StupidModel5 BugPopulation lt lt Scripting gt gt observer population maxFoodConsumption Double nbBug Integer 1 StupidModel5 Bug lt lt Scripting gt gt no Integer axFoodConsumption Double StupidModel5 Space lt lt Scripting gt gt StupidModelS Cell lt lt Scripting gt gt nbLine Integer maxFoodProduction Double nbCol Integer maxFoodProduction Doublel y Integer x Integer Figure 7 24 The stupid conceptual model 5 e the space has to do the same thing for its cells for i in range nbLine for j in range nbCol cell port cell i nbCol j addPort cell StupidMode15 Cel1 parameters pair x i pair y j pair maxFoodProduction maxFoodProduction lt HERE linkPort portRef cell space
104. post an internal event post a logical event wit post a logical event reply to an influence wi reply to an influence post a probe without az post a probe creates a port ref creates a pair for creates parameter creates a new ent links referenced p removes a referen destroys itself Appendix C The structure package The aim of the structure package is first to provide the new data structures suitable to represent the structural part of the model as well as eventually the information exchanged by the DEVS modules In the long run this generic data structure is assumed to be the target implementation described by an ontology Therefore we also developed an ontology language structure suitable to be mapped into this implementation For the time being it is used to provide useful data structures for implementing the DEVS module states The structure package is divided into four main boxes e The data box implements the data structure itself The management of dimensions units quantities and references is factored out for genericity e The meta data box implements the description of the data structures It is used for consistency checking in data structure manipulations e The assertion box abox implements the corresponding part of the ontol ogy language It is implemented to test its capability to fully generate an associated data structure e The terminological box tbox implements the corresponding part of the ontology lang
105. pter 5 In figure 4 5 one shows the attribute panel where the local attributes can be added or deleted through a popup menu Additionally one can see the list of inherited attributes as shown in figure 4 6 but this list cannot be edited Only the locally defined attributes can be edited the inherited list being computed 4 2 3 Deleting a category A category can be deleted by selecting it and selecting Remove from the Edit menu or by right clicking on it and selecting Remove in the popup menu It is asked whether the category must be removed from the ontology lFor UML literates it looks like a stereotype and in fact it has a related semantics with respect to the MDA specifications CHAPTER 4 THE CONCEPTUAL MODEL EDITOR 26 Name Abstract ma E Te Super category Farm Person 4 Documentation Attributes Relations Dynamics Attribute declarations Local attributes Inherited attributes O Attribute Cardinality Type cashFlow 1 Float Ok Cancel Figure 4 5 The category editor with the attribute panel Name Abstract EJ Super category Farm Person Documentation Attributes Relations Dynamics Attribute declarations Local attributes O Inherited attributes Attribute Cardinality Type age 1 Integer cashFlow 1 Float name 1 String Ok Cancel Figure 4 6 The category editor with the inherited attributes e if yes the category is removed both from the drawing and the list of categ
106. rete time e the rolling ball is submitted to uniform movement described by the fol lowing equations x t Lo Ve X t y t yo vy X t e at random time the kicker computes a random two dimensional vector lt kz ky gt which is sent to the ball to change its trajectory in the following way Ve Vz kg Vy Vy ky This example has been programmed in Java and the user could look at the source code of the package example where the classes Clock RollingBall Kicker and Observer have been defined To run it open the project example pml in the folder example The project contains two conceptual models Examplel is the conceptual model of the rolling ball example Example2 is the conceptual model of a cellular automata for simulating fire spread To the conceptual models correspond two concrete models defining an instan tiation the definition of some outputs and the associated control panel To actually run the simulation select the corresponding simulation model and suc cessively push on the buttons Reset Initialize and then Step as many time as you want or Run after having entered an end date 61 CHAPTER 7 SOME EXAMPLES 62 example RollingBall lt lt NativeState gt gt observer example Kicked lt lt NativeState gt gt kicked seed Zi bservedaxample Observer lt lt NativeState gt gt Figure 7 1 The conceptual model for a kicked and observed rolling ball 7 2 1 Def
107. rmat is imposed for the content A content is necessarily an array at the implementation level an instance of Java array of e arrays allowing recursive structures e simple types shorts integers longs floats doubles booleans and strings respectively implemented internally in Java as instances of Short Integer Long Float Double Boolean and String In principle no other kind of data can be send through the influences if one want to ensure consistency between entites writtent in different programming languages 5 2 4 The probes It is possible to associate to any individual therefore to any M DEVS entity a visualization window for displaying any information evolving over time e g the entity state changes Having no hypothesis on the nature of the entity states there is NO automatic synchronization between the model and its visualization To perform this visualization one has to declare a list of probes given by their name type only simple types are allowed and cardinality in reality it is just another type of influence When specifying the behavior i e the various transition functions the user has to explicitly send probe values whenever he wants to signal a change The probe value is propagated to the visualization window which can perform whatever one wants drawing or saving the data for further processing 5 2 5 The time The underlying time for the whole system is considered discrete regardless of the gr
108. rne Externe Confluent gt x0 self toDouble self getParameter x0 yO self toDouble self getParameter y0 vx self toDouble self getParameter vx vy self toDouble self getParameter vy self sendProbe state self list x0 yO vx vy System out printin Ball initialized x0 y0 vx vy Ok Cancel Figure 7 6 The rolling ball category with the initialize panel The user is asked to further explore the model which is available as an example to see how the behaviors are defined in the various scripting languages 7 2 3 Defining the concrete model As said before the definition of the structure and dynamics is part of the concep tual model and cannot be run directly From the conceptual model a concrete and simulatable model can be instantiated You have to open a concrete model editor At the top of the right panel you have a list of conceptual models you can take your definitions from The figure 7 8 shows a window in which a model has been built by creating an instance of each of the categories an instance of clock has been added to define the time rate at which the observer samples the rolling ball In this figure each port is linked to the proper entity The drawing panel uses a modified UML object diagram The links are named which is not the case in UML As in UML the name of the instances is optional and for documentation purpose only The actual structure of an individual is not only compose
109. rt void sendExternal String portName String influenceTypeName void sendExternal Port portName String influenceTypeName void sendExternal String portName String influenceTypeName Object args void sendExternal Port portName String influenceTypeName Object args void sendLogical String portName String influenceTypeName void sendLogical Port portName String influenceTypeName void sendLogical String portName String influenceTypeName Object args void sendLogical Port portName String influenceTypeName Object args void sendInternal int duration String influenceTypeName void sendInternal int duration String influenceTypeName Object args void reply LogicalInfluence influence String influenceTypeName void reply LogicalInfluence influence String influenceTypeName Object args These methods can be called in most methods e to signal a state change by a probe public void sendProbe String name Object args e to destroy itself CHAPTER 5 THE BEHAVIOR 41 public void die It removes the entity from the scheduler removes of the link references as well as all the scheduled incoming influences e and finally to stop the entire simulation public void stop It is used to be able to stop the simulation whenever some conditions occur In addition a number of methods are defined to dynamically create and link entities during the simulation e v
110. simplified version of which documentation can be found in 1 is used The following section provides the primitives for Mimosa B 1 Mimosa primitives In general we defined the class Self with the class methods which are defined in this section For Mimosa we added two methods to create and access local variables it is guaranteed to have a copy of these for each entity Self at lt var gt put lt value gt to create or change a local variable Some functions are provided to access the Mimosa random generator newRandom creates a random generator newRandom lt seed gt creates a random generator nextBoolean lt random gt generates a boolean randomly nextInt lt random gt in lt n gt generates an integer from 0 to n nextDouble lt random gt generates a real from 0 to 1 Finally the access to the DEVS entity functionalities are provided as follows e for each parameter the variable with the same name is defined with the associated value within the local variables It can be accessed through the method Self lt parameter gt e when a script for a DEVS function is called the local variable time is linked to the duration elapsed since the last internal or external transition and can be accessed by Self lt time gt e each influence is a Java object whose structure can be accessed by the following functions is lt influence gt type lt name gt True if the influence has the given name contentOf lt influ
111. subpart of M DEVS We have provided two versions corresponding to most needs e StaticObjectis used when the only functionality is around state variable values being set and get e SimpleObject is an extension of StaticObject where external and logical influences are considered as method calls the external influences when the SimpleObject will change state in response and the logical influences when only information updates and requests have to be handled StaticObject contains a set of state variables to choose among the at tributes The following incoming influences are expected e setState name value as an external influence to change the value of one of the variables e getState name as a logical influence to ask for the value of one of the variables The following outgoing influences are issued in response to the getState influ ence e state name value as a logical influence to communicate the value of the requested state variable e undefinedState name as a logical influence to communicate the state variable has no value SimpleObject has the same semantics as StaticObject and as such pro vides to the same incoming and outgoing influences In addition to defining the state variables the modeler can add as many additional incoming and outgoing influences as he wants SimpleObject allows to associate a piece of code to execute to each incoming influence In figure 5 3 the upper part shows on the left the list of defin
112. t has key using equal and returns the found pair if any f otherwise A 5 Mimosa primitives For Mimosa we added three very common control structures for better read ability when lt cond gt lt expj gt lt exp gt executes if t unless lt cond gt lt exp1 gt lt expn gt executes if f for lt var gt lt list gt lt exp1 gt lt exp gt a simple loop over a list times lt var gt lt nb gt lt exp gt lt exp gt a simpler loop repeated nb times Some functions are provided to access the Mimosa random generator newRandom lt seed gt creates a random generator nextBoolean lt random gt generates a boolean randomly nextInt lt random gt lt n gt generates an integer from 0 to n nextDouble lt random gt generates a real from 0 to 1 Finally the access to the DEVS entity functionalities are provided as follows e the variable self is linked to the current Java state e for each parameter the variable with the same name is defined with the associated value within the global context It can additionally be accessed through the function getParameter lt sym gt e when a script for a DEVS function is called the global variable time is linked to the duration elapsed since the last internal or external transition APPENDIX A INTRODUCTION TO SCHEME 100 e each influence is a Java object whose structure can be accessed by the following functions is lt influence
113. t MAL Re 29 5 2 The operational semantics 30 D2511 Fhe model 2 56 2 esac kee eee oa a Rares 31 0227 Th ports 4408 444 dete oS SU Rte ele book A 33 5 2 3 The influences 33 52A Th probes ti aa ek a 34 52 0 Letelier does nos ae 2 et es GEE 34 CONTENTS 5 3 6 The 6 1 6 2 6 3 6 4 The behavior specification 5 3 1 Programmatic specification 5 3 2 Scripted specification 5 3 9 State Charts A ee RU EN 4 5 3 4 Further extensions concrete model editor Individual edition 6 1 1 Drawing an individual 6 1 2 Editing an individual 6 1 3 Deleting an individual Link edition eus e see get oe fran at Soir e RS 6 2 1 Drawing a lnk vis sata nt dot RG 6 2 2 Deletingalink Output specification 6 3 1 Drawing an output 6 3 2 Editing an output 6 3 3 Deleting an output 6 3 4 Drawing an output edge 6 3 5 Deleting an output edge Control panel definition 6 4 1 Drawing an output view 6 4 2 Deleting an output view 6 43 Drawing a parameter editor 6 4 4 Deleti
114. tes and of course a part of a space Position is a mapping from the bugs into the cells giving the individual posi tion of each bug It also encodes the topology of the space by defining the cells neighborhood We provide a seed attribute for its random generator The resulting conceptual model is shown in the figure 7 11 The position refer ences both the bug population and the space Each bug knows the position to be able to position itself and move From the dynamical point of view we shall describe in turn each of the category e the bug population has only the function to create the bug population Therefor we define the initialize function as follows for i in range nbBug p port bug i addPort p StupidModel1 Bug parameters pair no i linkPort portRef p position portRef observer sendProbe population nbBug sendLogical observer populationSize nbBug The first four lines are used to generate the bug population The last two lines informs the position respectively any probe observer of the popula tion size e the space is doing exactly the same for the cells CHAPTER 7 SOME EXAMPLES 69 1 StupidModel1 BugPosition lt lt Scripting gt gt l space observer StupidModel1 Space lt lt Scripting gt gt nbLine Integer nbCol Integer cell StupidModel1 Cell lt lt Scripting gt gt x Integer Figure 7 11 The stupid conceptual model 1
115. the XML file shall hopefully be presented in the programmer s manual 5 3 2 Scripted specification The previous procedure being relatively heavy but necessary if one wants either an efficient piece of code or to use Java to encapsulate a legacy simulation software we provide the same functionality by using scripting languages directly through the user interface The basic principles are the same and we are using the same names for the variables and functions or equivalent for consistency For using this functionality you have to select Scripting in the drop down menu of the behavior pane In the state panel you will have another drop down menu to select the desired scripting language as well as an editor for specifying the structure of the state In a model any combination of scripting languages can be used because all the specific data structures are translated into a standard Java format and back to the specific data structures So feel free to use any one you find most appropriate for your usage Of course it requires to be multi lingual Java scripting Java scripting makes available the full Java language by using the bean shell library see 5 for getting the related documentation In particular all the methods defined in the section 5 3 1 are readily available However to call them a new variable is defined self The methods can be called by addressing them to self For example for the Ae function the code is CHAPTER 5 T
116. the right and the probes or observations on S from the top The function init is called for initializing the entity when created either at the beginning of a simulation or when created The dynamics is described by the following functions deat is a function to specify the reaction to a set of incoming influences all the influences occurring at the same time are given simultaneously CHAPTER 5 THE BEHAVIOR 32 dint is a function to specify the internal change when and what to do is spec ified by Aint con is a function to specify the reaction to the simultaneous occurrence of an internal change and the arrival of a set of incoming influences dog is a function to specify the reaction to a set of logical influences possibly producing further logical influences Xext is a function to provide the outgoing influences when it is called is also specified by Aint Aint is a function to provide the what to do an internal influence and when called 7 the duration to wait before doing the internal change Alog is a function to provide the logical influences to occur after each transition Astr is a function to provide the structural changes to occur also after each transition Aprobe is a function to provide the observables of the state changes after each cycle For all functions but init the duration since the last cycle see below is given as an argument Therefore the internal logic of any atomic model is bas
117. thout the associated conceptual models CHAPTER 2 RUNNING MIMOSA 12 Ontologie example example xml DefaultOntology Graphe Cat gories Types d influences example Le m S Sa y Panneau clocked example Clock f example EntityType lt lt NativeState gt gt lt lt NativeState gt gt example JavaClock example SchemeClock example PythonClock example HardClock lt lt LanguageState gt gt lt lt LanguageState gt gt lt lt LanguageState gt gt lt lt Clock gt gt example RollingBall example Kicked lt lt NativeState gt gt observer lt lt Nativestate gt gt kicked bserved xample Observer F yo lt lt NativeState gt gt seed is w E example HardKicker example Javakicker example SchemeKicker lexample PythonKicke example HardObserver exampleJavaObserver ex __ lt lt Kicker gt gt lt lt anguageState gt gt lt lt LanguageState gt gt lt lt Agent gt gt lt lt Observer gt gt lt lt LanguageState gt gt lt seed example HardRollingBali example JavaRollingBali example SchemeRollingBall example PythonRollir lt lt RollingBall gt gt lt lt LanguageState gt gt lt lt LanguageState gt gt lt lt LanguageState gt Figure 2 2 The conceptual model editor as an example of an editor window Help this men
118. top CHAPTER 7 SOME EXAMPLES 89 Bo Bug size distribution nbBug 15 amp 3 10 1 2 maxFoodConsumption e ow maxFoodProduction M Show grid C Spread objects Zoom out Record image Figure 7 31 The control panel with the histogram Running the simulation model The simulation shall stop whenever the end date or the condition is met either case comes first If one want the simulation to stop only when the condition is met one must provide a very large end date It is thought as being a security in case the condition is never met 7 3 8 Stupid model 8 This version illustrates the possibility to output information to a file In this case it is desired to output the minimum mean and maximum bug size at each cycle Defining the conceptual model The conceptual model structure does not change and remains the one of the figure 7 24 The only difference is the computation of the desired values e all the necessary information is already recorded in the bug population for the histogram It is easy to also compute the min max and mean from the available information each time the sizes change in 0 7 the management of the histogram has been removed for readability for inf in getAllInfluences if type inf size CHAPTER 7 SOME EXAMPLES 90 ProbeOutput out12 CategoryChartView out11 null StupidModel8 BugPosition lt lt Defaultindividualinitializer gt gt Discrete2DSp
119. trated in the figure 6 11 This dialog is composed of two parts e the upper part lists all the parameter editors available in the selected control panel Selecting one of these and typing either return or pushing the Existing button shall draw the corresponding parameter editor at the selected place e the down part is used to create a new parameter editor with two drop down menus the first one is for selecting one of the individuals created in the concrete model panel see 6 1 1 the second one is for selecting one of the attribute to edit of the individual 6 4 4 Deleting a parameter editor A parameter editor can be deleted by selecting it and selecting Remove from the Edit menu or by right clicking on it and selecting Remove in the popup menu It is asked whether the parameter editor must be removed from the model e if yes the parameter editor is removed both from the drawing and the list of parameter editors defined for the model e otherwise only the drawing is removed but the parameter editor remains unchanged Chapter 7 Some examples 7 1 Introduction This chapter is considered as a stepwise tutorial to develop your own models All these examples are provided in the distribution in the example folder 7 2 The rolling ball example As an example we shall model a simple system composed of one rolling ball and a kicker This example allows the illustration of a combination of continuous and disc
120. u gives access to a number of tools for debugging Statistics displays in the output window some statistics about the data structures used by the scheduler number of created entities and usage of the influences Predefinitions displays in the output window the predefinitions as de fined in the scripting mechanism Show content displays in the output window the content of the tables created by the various editors which are the data structures behind the scenes Script interpreter displays a window in which the user can enter ex pressions in any of the provided scripting languages in order to test the code The results are displayed in the output window when push ing the eval button 2 4 The editor windows Each editor window has the same structure see the figure 2 2 It is divided in two vertical panels The left panel contains the list of existing models either conceptual or con crete referenced by their names or the projects hierarchies In Mimosa these models are also referred to as ontologies One can select an existing concep tual model in the conceptual model editor or concrete model in the concrete CHAPTER 2 RUNNING MIMOSA 13 Ontologie example example xml DefaultOntology Graphe Cat gories Types d influences example SchemeObserver PythonKicker JavaClock Kicked SchemeClock Observer JavaRollingBall EntityType HardKicker JavaObserver RollingBall JavaKicker HardClock Clock Scheme
121. uage It is implemented to test its capability to derive a corresponding abox in the one end and the meta data structure on the other end C 1 The meta data box The meta data box in the package structure metabox is described first be cause it defines the structure of the data box and in particular constrains the set of possible value structures T his box introduces the notion of Classifier A classifier is the specification of a set of possible value structures In particu lar it defines a test whether a value is in the given set For the time being the classifiers are defined into four categories e the types to describe the data structures within hierarchies of descriptions e the dimensions to describe the quantities 103 APPENDIX C THE STRUCTURE PACKAGE 104 e the tests to describe possible values in intension e the values to describe themselves A classifier defines only one important method check Value value that returns true if the value is member of the set spec ified by the classifier false otherwise C 1 1 The types A type is the representation of a set of values The types are organized into hierarchies specifying set inclusions For example LivingBeing is a type de scribing the set of all living beings Animal is the set of all the animals which happens to be a subset of LivingBeing LivingBeing is said to be the super type of Animal and conversely Animal is said to be a subtype of LivingBeing A t
122. ulationSize nbBug Each time the sizes of the bug sizes are signaled through logical influences the histogram is updated in dog for inf in getAllInfluences if type inf size id size contentOf inf bugSizes id size sendProbe objectState bug id size size for i in range len sizeHisto sizeHisto i 0 for size in bugSizes if size gt 10 sizeHisto 9 1 else sizeHisto int size 1 for i v in enumerate sizeHisto sendProbe category i size v Nothing else is needed Notice that the systematic use of aggregates like the bug population and the space simplifies dramatically how to compute aggre gated values as the histogram A similar computation could be done for the distribution of food availability Defining the concrete model In the concrete model one must add a new visualizer for the histogram and link it to the bug population instance The result is illustrated in the figure 7 28 The histogram visualizer is described in the section 6 3 2 and can be parameterized as shown in the figure 7 25 The control panel is now changed to add three new parameters editor see 7 31 In effect any attribute of the concrete model can be associated to an editor to be put on the control panel if needed CHAPTER 7 SOME EXAMPLES 87 CategoryChartView out null StupidModel6 BugPosition lt lt Defaultindividualinitializer gt gt seed 5612653141 population space ll
123. ultiplication by an element of Z Any vector space S of dimensions is implemented as an instance of the class DimensionSpace with the following methods add String symbol creates and defines a new magnitude dimension noted with the given symbol add Type type creates and defines a new multitude dimension defined as a number of instances of the given type getDimensions to get the ordered list of atomic dimensions getSize to get the number of defined dimensions isCompatible DimensionSpace dimensionSpace true if the two dimension space are compatible A dimension space will be the reference for all the dimensions defined as a combination of these atomic dimensions In particular it is not possible to combine two dimensions from two incompatible dimension spaces The DimensionSpace class defines the following constant e ONE as the dimensionless dimension with no symbol as well as the following constructors DimensionSpace DimensionSpace dimensionSpaces that creates and defines a new dimension space as a combination possibly empty of ex isting dimension spaces The possibility to build dimension spaces from others allows to factorize the definitions that are very often used although not necessarily standardized In particular it allows to build on the following predefined dimension spaces e the physical dimension space defines once and for all the seven standard physical dimensions e the
124. uments shall be saved in the file one per line optionally preceded by its name and its global and local time stamps Plots A number of probe observers for plotting data has been introduced e the category chart view is plotting series of which values are given by categories e the series chart view is plotting series given by x y coordinates e the times chart view is plotting series against the probe time stamps More precisely a category chart view is a probe observer that visualizes series plotted by categories It is parameterized by e the chart name e the x axis name the y axis name the type of chart e whether it has a legend or not e whether it has tool tips or not It understands probes of the form category lt categoryName gt lt seriesName gt lt value gt where the names can be any string and the value is expected to be a double or convertible to a double The name of the series is optional when there is only one series to display A series chart view is a probe observer that visualizes series given by x y pairs It is parameterized by e the chart name e the x axis name the y axis name the type of chart e whether it has a legend or not e whether it has tool tips or not It understands probes of the form series lt seriesName gt lt xValue gt lt yValue gt where the name can be any string and the values are expected to be doubles or convertible to doubles The name o
125. values are integers The object type is optional if there is only one object type and so it is for the layers GraphView Coming soon 6 3 3 Deleting an output An output can be deleted by selecting it and selecting Remove from the Edit menu or by right clicking on it and selecting Remove in the popup menu It is asked whether the output must be removed from the model e if yes the output is removed both from the drawing and the list of existing outputs defined in the model e otherwise only the drawing is removed but the output remains as an existing output 6 3 4 Drawing an output edge To draw an output edge in a given place it is enough to click on the corresponding button and then from an individual called the source individual to an output called the target output or to right click at the desired place to show up the same toolbar as a popu menu new dialog is opened as illustrated in the figure 6 11 This dialog is composed of the list of available output edges between the individual and the output 6 3 5 Deleting an output edge An output edge can be deleted by selecting it and selecting Remove from the Edit menu or by right clicking on it and selecting Remove in the popup menu It is asked whether the output edge must be removed from the model CHAPTER 6 THE CONCRETE MODEL EDITOR 58 CD gt D o fae Figure 6 12 The control panel toolbar e if yes the output edge is removed both from t
126. wnership 0 surface 10 0 age 30 cashFlow 1200 0 name Parker p2 Farm Plot Paul Farm Farmer surface 10 0 ip 0 age 45 ownership 1 cashFlow 5300 0 name McGregor p3 Farm Plot e a surface 15 0 Figure 6 6 The example of links 6 2 2 Deleting a link A link can be deleted by selecting it and selecting Remove from the Edit menu or by right clicking on it and selecting Remove in the popup menu It is asked whether the link must be removed from the model e if yes the link is removed both from the drawing and the list of links defined for the model e otherwise only the drawing is removed but the link remains unchanged 6 3 Output specification The figure 6 7 shows a concrete model with three individuals and one output The arrows are connecting the individuals to an output which is in this case a 2D grid view specifying that the corresponding probes must be sent to that output 6 3 1 Drawing an output To draw an output in a given place it is enough to click on the corresponding button and then at the desired place or to right click at the desired place to show up the same toolbar as a popu menu new dialog is opened as illustrated in the figure 6 8 This dialog is composed of two parts e the upper part lists all the outputs available in the selected model Select ing one of these and typing either return or pushing the Existing button shall draw the corresponding output at the selected place
127. wth is random within a given range maxFoodProduc tion The spontaneous growth is managed by the usual functions Aint sendInternal 1 grow CHAPTER 7 SOME EXAMPLES 78 and dint availableFood availableFood maxFoodProduction nextDouble rand sendLogical space food x y availableFood The reaction to the consumption request is naturally managed by deze for inf in getAllInfluences if type inf consume request contentOf inf consumption min request availableFood sendLogical bug consumed consumption availableFood availableFood consumption sendLogical space food x y availableFood where the consumption depends on both the request and the available food However it could happen that the spontaneous growth and the consumption occurs at the same time provoking a confluent transition deat It combines both 6 n and dex and decides in which order things happen in this case growth is before consumption for inf in getAllInfluences if type inf consume availableFood availableFood maxFoodProduction nextDouble rand request contentOf inf consumption min request availableFood sendLogical bug consumed consumption availableFood availableFood consumption sendLogical space food x y availableFood e the position incurs only one change changing the position of a bug implies to modify the reciprocal links between the bug and the c
128. x y availableFood sendProbe food x y availableFood lt HERE but also in ert dine and dcon CHAPTER 7 SOME EXAMPLES 82 cell 11 9 ProbeView Add Cancel Figure 7 23 Choosing a probe observer for cells and bugs Defining the concrete model The concrete model remains strictly the same Running the simulation model The figure 7 23 shows dialog which appears when clicking on a cell There are two drop down menus e the top one is a list containing the clicked cell if any there is no cell in a continuous space and the objects in the cell or around the click position e the bottom is a list of available probe observers for monitoring the chosen object Clicking on the add button creates and connects the chosen probe observer 7 3 5 Stupid model 5 This version still does not add any new behavior but provides the possibility to parameterize the simulation by the initial number of bugs the maximum daily food consumption of the bugs and the maximum food production of the cells Defining the conceptual model The conceptual model structure changes slightly to have the maximum food consumption and food production appear as attributes in the relevant categories see 7 24 Now each cell and bug has the maximum food consumption and production as attributes as well as the bug population and space because these two last categories are in charge of initializing the bugs and cells res
129. xpress any computation one could dream of Nevertheless the resulting syntax would become unreadable for any reasonable computation The simplest way to overcome this problem is to provide the possibility to associate names to expressions with the form define lt name gt lt expressions gt number of names have been predefined in Scheme for all the current arithmetic operations as well as the operations on very common data structures By the way define is not a function name but the name of a syntactic form which is transformed behind the scene in a proper application The set of possible syntactic forms can itself be extended parameterizing the Scheme interpreter with high level constructs at will not explained in this introduction A structure or object is also called a literal expression is of the form quote lt something gt or alternatively lt something gt The something is either e a number t and HH a character e a string a symbol e apair lt something gt lt something gt or a list lt something gt lt something gt a vector tt lt something gt lt something gt 96 APPENDIX A INTRODUCTION TO SCHEME 97 The first four categories do not need the quote because they self evaluate i e their value is themselves Finally an additional power is acquired by the relationships between struc tures or objects and expressions Of course expressions transform structures into
130. ype does not need to have a name allowing for anonymous types Therefore we have two constructors Type for anonymous types Type String name for named types Used this way having the same name does not mean it is the same type We use name spaces to guarantee this property see section A type is an instance of the class Type with the following methods addSubtype Type type adds a subtype to a type setSupertype Type type sets the super type of a type a type can only have one super type removeSubtype String symbol removes a subtype from a type isSubtype0f Type type checks whether it is a subtype of the given type isSupertype0f Type type checks whether it is a supertype of the given type C 1 2 The dimensions The quantities are anything that result from a measure or a count More than one feature can be measured or counted on an object For example one can measure the weight or the height of a tree or the number of apples in it The feature that can be measured or counted is called the dimension More specifi cally something is measured along a dimension called a magnitude or counted along a dimension called a multitude Therefore the length the weight or the number of apples are dimensions and any quantity is characterized by the di mension it measures Weights and lengths are magnitudes while the number of apple is a multitude The International System of Measure defines seven atomic dimensions

MIMOSA user's manual (Draft version 1.4.0) - UR Green

Contents

Download Pdf Manuals

Related Search

Related Contents