an experimental collective intelligence research tool
Contents
1. Figure 27 our simulation result Comparing our simulation results figure 27 with that of the Wasp Nest Building Simulator it can be seen that both simulations have resulted in nature like patterns 1 19 Bonabeau believes that extensive simulations on a powerful computer have to be performed in order to explore the behavioural space in a satisfactory manner even in this simplest case 1 Our software proves that small scale simulation runs smoothly and relatively fast even on a PC We offer 3D rotations during simulation enable users to observe the visualization from various angles While Wasp Nest Building Simulator 1s implemented using C C and is to be distributed under GNU License soon our software is solely implemented in Java which is platform independent Furthermore our implementation integrates neatly with the Repast toolkit framework and may provide future researchers convenience and user friendly interface Video clips showing the actual software and hardware simulations are available upon request by contacting the authors As a conclusion for the software simulation as indicated by Bonabeau et al we have restricted our attention to very simple individual algorithms where information 1s processed locally in space as well as in time 1 We hope the simulation results which closely resemble these found in 1 can be of some value to the understanding of biological collective systems 1 Though2. Our software implementation observes and predicts possible outcomes by defining a number of termite and environmental variables such as the density of obstacles wood chips Figure 5 is a screen shot for the software simulation in action The red rectangle represents termite carrying no woodchip orange rectangle represents termite carrying one woodchip yellow rectangle represents woodchip During the simulation the user observes the movement of red and orange termites picking up or dropping woodchips in the simulation space je j i Fig A i a Tack tunt Pe a DE F Figure 5 2D termite simulation In order to visualize the 3D colony we use Java 3D API Java 3D defines the concept of a virtual universe as a three dimensional space with an associated set of objects 7 Since Repast doesn t come with 3D visualization support we separately programmed a set of Java classes to be integrated with Repast to realize and illustrate the 3D colony architecture Our 3D integration works with the Repast original GUI in a way that the display surface corresponds to commands sent from various buttons such as setup step pause and stop With proper time delay between each clock tick the users can observe the growth of the artificial colony architecture Furthermore we incorporate mouse rotation functionality into the 3D visualization At each clock tick users are able to rotate the visual colony by left mouse click so tha
3. sensor and right sensor of the gripper figure 14 Second if the left senor signal is on meaning that the robot detects an object from the left we let the robot turn left until the right sensor turns on This indicates that the robot has just passed the object figure 15 Thus we will let the robot turn right for a little an angle of around 3 degrees to centre the object into the gripper figure 16 Now the robot can keep moving straight until the close sensor is on this means the object is inside the robot and grips the object figure 17 Afterwards the robot starts searching for a new object When it hits the new object by its whisker it releases the object it is carrying After releasing the object the robot moves backward turns an angle of 45 degrees and the same procedure 1s repeated Figure 14 Figure 16 Figure 17 The robot can detect a new object by using its left and right sensors We use a touch sensor to enable the detection of the second object in order to release the one it carries However there is one drawback if the new object the robot encounters is too big it could activate the touch sensor and the robot would release the object right after gripping it 5 RESULTS AND CONCLUSIONS Simulation results are shown below as screen shots for the software implementation Figure 18 19 and 20 show 2D termite simulation hon Termites Scape Options Be aR w Figure 18 initial stage Termit
4. AN EXPERIMENTAL COLLECTIVE INTELLIGENCE RESEARCH TOOL Bei Wang Dung Hoang Idris Daiz Chiedu Okpala and Tarek M Sobh Department of Computer Science University of Bridgeport Bridgeport CT 06601U S A Abstract The Collective Intelligence Research Tool CIRT is an experimental software and hardware research tool It provides an inexpensive and efficient alternative research implementation that demonstrates simulations of the collective behaviour of self organized systems primarily social insects The software focuses on 2D simulations of the woodchip collecting behaviour of termites and 3D simulations of the building behaviour of wasps The hardware simulation employs a Boe Bot robot which has the potential of simulating simple movements of a social insect by extending its functionality through adding sensors and integrating a control chip Keywords Artificial Life Intelligent Agents and Multi Agent Systems 1 INTRODUCTION Social insects are known to be capable of producing complicated colony patterns 1 Our first project objective is to simulate self organized systems using social robots We have implemented a robotic termite agent which is able to simulate the wood chip collecting behaviour of the termite By defining the behaviour for one robotic agent we could potentially observe the collective building activity of a group of robots From a software viewpoint our goal 1s to simulate and visualize the collective bui
5. architecture 1 For example when there 1s no brick in the central cell a wasp puts down a brick of type 1 in the case of configuration 1 in figure 1 9 cells above are all already filled with type 2 bricks and type 2 in the case of figure 2 4 1 There are 9 configurations in total Detailed rules can be found in 1 Furthermore taking symmetries into account each rule expands further For example configuration in figure 2 expands to more configurations as shown in figure 3 and 4 SOO CI I I Z Fo me haga aa 1 T a Q 2 m Figure 4 Figure 3 Figure 1 4 Local neighbourhood in 3D lattice swarm 1 3 SOFTWARE IMPLEMENTATION Our simulation uses the Repast framework an agent based modelling toolkit for java It has three major classes agent space and model Employing the Repast software architecture an agent class describes how an agent interacts with the environment and moves around the space 1 A model class coordinates the setup and running of the model A space class defines the environment such as the distribution of woodchips for the termites 2D simulation and the coordinates of wasps and bricks in swarm 3D simulations 10
6. es Scape l Options 10 x Figure 19 in progress a Termites Scape Options a O x Figure 20 final stage Figure 21 shows a lattice swarm simulation done within a 20X20X20 3D space with 315 tick counts note there is an X Y axis displayed in this simulation 10 x Custom Actions Repast Actions l x Parameters amp Termites Scape Model Parameters Options NumAgents 10 Size 19 RePast Parameters CellDepth CellHeight pees Pauseat RandomSeed 1059884778880 seeseeceeescscscececesescescescscccece SEE He ee eee ee ee ee ee ee ee ee My ey o00oo0no0n0nom0nm0n0onn0no0n0n0A SRISARINK ao ao aa aa a a o ao a a No ao a co oo oop A Figure 21 Figure 22 25 shows a lattice swarm simulation done within a 20X20X20 3D space with 38142 tick counts Figure 26 shows a simulation done within a 40X40X40 3D space with 91306 tick counts Due to the huge size of the simulation space the resulting structure remains relatively small scale even after a large number of tick counts Figure 24 j 1ol x Q EJ Tick Count 91306 0 i xi Custom Actions Repast Actions f Parameters RePast Parame ters CellDepth 5 CellHeight 5 Celidth 5 RandomSeed 1060001131092 Figure 26
7. f construction each swarm insect automatically responds dropping bricks when it meets any local configuration As explained by Bonabeau et al the regulation of the building activity is mainly achieved by the nest structure instead of depending on the workers themselves 1 The wasps put bricks into a 3D structure with the following behaviours 1 1 Each wasp is born at a random location in the 3D space 2 The wasp observes its local configuration with 26 neighbouring cells 3 Ifthe local configuration applies to one of the pre defined patterns the wasp drops a corresponding brick at that location and then moves to another random location 4 Ifthe local configuration doesn t apply to any of the patterns the wasp does nothing and moves to another random location 5 The result of the construction eventually produces certain architectures that can be found in nature CIRT simulates and visualizes in 3D space the growth of the colony Social wasps act in 3D space and drops bricks based on pre defined behaviour rules The software observes the outcome by redefining the number of wasps According to Bonabeau s research the neighbourhood of the wasp is composed of the 26 cells surrounding the central cell it occupies 1 This neighbourhood consists of 3 3X3 layers along the y axis see figure 1 1 When the wasp occupies the central position of the layer y marked in black it follows certain rule to produce our 3D
8. ftware and integrate it into Repast to enhance its simulation possibilities REFERENCES 1 Bonabeau E Th raulaz G Arpin E and Sardet E The building behavior of lattice swarms In Artificial Life IV Brooks R and Maes P eds pp 307 312 MIT Press 1994 2 Ramos V On the Implicit and on the Artificial Morphogenesis and Emergent Aesthetics in Autonomous Collective Systems In ARCHITOPIA Book Catalogue Art Architecture and Science J L Maubant and L Moura Eds pp 25 57 Minist rio da Ci ncia e Tecnologia Feb 2002 3 Langham A E and Grant P W Evolving the Building Activity of a Termite Colony for Finite Element Mesh Generation Department of Computer Science University of Wales Swansea Research Report CSR1 4 99 1999 4 A Cooperative Multi Robot Control Architecture Dynamic Concepts Inc Technical Report 2002 5 Fong T Nourbakhsh I and Dautenhahn K A Survey of Socially Interactive Robots In Robotics and Autonomous Systems vol 42 3 4 March 2003 6 Krieger M J Billeter J B Keller L 2000 Ant like task allocation and recruitment in co operative robots In Nature 406 992 995 2000 7 Bonebeau E Dirigo M and Theraulaz G Inspiration for Optimization from Social Insect Behavior In Nature 406 39 42 2000 8 Bonebeau E Theraulazx G and Cogne F The Design of Complex Architectures by Simple Agents In Sante Fe Institute Working Paper 98 01 005 9 Re
9. ks around randomly in the sample space 2 An empty handed termite picks up a randomly distributed wood chip if it comes across one The termite continues to walk around randomly 4 When the termite comes across another wood chip it finds a nearby empty space and puts its wood chip down and becomes empty handed again 10 oS CIRT simulates and visualizes in a 2D space the termites gathering wood chips into piles based on the initial behaviour definition It also observes and predicts possible outcome by redefining the number of termite and environmental variables such as the woodchip density As the simulation progresses the randomly distributed woodchips would end up in a single large pile as shown in similar simulations 10 2 2 3D Lattice Swarms Simulation The architectural patterns grown by artificial agents moving and acting in a virtual space in our case artificial wasps are based on biological data provided by observations of nests built by social wasps 1 We based our development on the simulation of the collective building of 3D wasps colony which involves two major objects wasps and their bricks According to Eric Bonabeau et al s research using stigmergic algorithms these agents move and act in a 3D lattice and are able to deposit bricks according to their local neighbourhood configurations 26 neighbouring cells for 3D lattice swarms using a look up table 1 In the stigmertic mode o
10. lding of complex architectures for termites in 2D space and social wasps in 3D space In addition to simulating self organized systems by changing variables such as the population and obstacle density the software provides an artificial life environment for observation of the emergent behaviour of autonomous agents in our case termites and wasps Current research on simulation of self organized systems and swarm intelligence have a shared underlying idea that the key feature of all nature s patterns is that they are self organized there is no guiding hand 1 9 Existing research projects include StarLogo StarLogoT NASA COIN project Repast AgentSheets Ascape and SWARM 10 17 StarLogo is a programmable modelling environment for exploring the workings of decentralized systems such as bird flocks traffic jams and market economies 10 RePast is a software framework for creating agent based simulations which provides a library of Java classes for creating and running agent based simulations 11 SWARM is a Contact author beiwang bridgeport edu software package for multi agent simulation of complex systems originally developed at the Santa Fe Institute 20 We have implemented the simulation of collective intelligence systems from both software and hardware perspectives as a complete experimental experience The 2D simulation of termites behaviour employs methodology found in the Starlogo project demon
11. object it is holding if there is one Here pin 4 and pin 6 connected to each switch circuit monitor the voltage at the 10 kQ pull up resistor When a given whisker 1s not pressed the voltage at the pin connected to that whisker is 5 V logic 1 When a whisker is pressed the I O line is shorted to ground and the pin sees 0 V logic 0 See figure 10 for details A program will keep checking whether the logic from pin 4 and pin 6 is changed If there is a change a corresponding subroutine will be called to react to the change by either releasing the object it carries or avoiding the object it touches 4 3 2 Gripper Module left sensor right sensor Figure 11 gripper The gripper figure 11 has 3 pairs of IR sensors used for object detection They are used to control the movement of the termite as well as the gripper The IR unit incorporates a standard IR LED with a 40 kHz IR receiver The IR specification as well as it schematic 1s given below figures 12 and 13 Size Width 15 8 mm Length 18 2 mm Power Requirements 5vdc 2 6 mA Vdd 5vdce Signal I O pin 0 J Signal vdd Ground Figure 12 IR specification Figure 13 IR schematic Note that the IR LED emitter and IR Detector are both connected to the same I O pin 4 3 3 Robotic Termite Working Scenario The termite works as follows first we let the robot spin left 360 degrees and keep detecting the signals sent by both the left
12. our software so far 1s modelled after the research done by Bonabeau et al it provides an inexpensive tool and a fast and platform independent implementation for researchers seeking new ways to implement these systems 6 FUTURE DEVELOPMENT Where can we go from here From a software perspective as Repast is an open source software framework for creating agent based simulations using the Java programming language and it is a more sophisticated development tool further developments might include developing a whole 3D visualization library that can be integrated into Repast for realizing powerful simulations From hardware perspective we might be able to get more robots to form a group of termite agents for better simulation results Meanwhile we may also install an extra pair of sensors for each robot to detect new objects Moreover it is also possible to bring in interaction between simulation software and hardware such as programming robotic agents behaviors through a software terminal The experimental robots may also lead to further research in the area of cataglyphis fortis 20 21 In addition biologists are concerned with the individual behavioral algorithms that allow a society to build its nest 1 Researchers have looked into genetic algorithms to implement nest construction algorithms 16 Therefore it is possible to implement an algorithm generation component or more precisely a rule generator for our so
13. snick M Turtles Termites and Traffic Jams Explorations in Massively Parallel Microworlds Cambridge Ma MIT Press 1994 10 StarLogo http education mit edu starlogo 11 Collier N Repast An extensible framework for agent simulation 2002 http repast sourceforge net 12 NASA COIN Project http s arc nasa gov AR projects ColInt html 13 SWARM http www swarm org 14 StarLogoT http ccl sesp northwestern edu cm starlogoT 15 NetLogo http ccl sesp northwestern edu netlogo 16 AgentSheets http agentsheets com 17 Ascape http www brook edu es dynamics models ascape 18 Boe Bot Specifications http www parallax com html_pages robotics boebot 19 Wasp Nest Building Simulator http www iasc enst bretagne fr PROJECTS SWARM nest html 20 Roumeliotis S I Pirjanian P and Mataric M J Ant Inspired Navigation in Unknown Environments In Proc 2000 AAAI International Conference on Autonomous Agents Barcelona Spain June 3 7 pp 25 26 21 Kuipers B and Byun Y T A robot exploration and mapping strategy based on semantic hierarchy of spatial representations In Robotic and Autonomous Systems 8 1 2 47 63 Nov 1991
14. stration and biological observations 10 3D simulation in our project focused on the building behaviour of social wasps using the methodology found in the work of Eric Bonabeau et al 1 Our hardware simulation draws idea from research done by Krieger M J 6 given robots with the ability to perform simple object removal tasks researchers are able to simulate collective behaviour among cooperative robots in our case termite agents 6 2 SOFTWARE SPECIFICATIONS Our simulation software is built around the Repast framework The software adopts the Repast graphic user interface GUI The Repast GUI 1s able to initialize start pause and stop a simulation It also enables user to alter some of the simulation variables such as the size of the display surface sample space and number of agents 10 Repast is able to handle 2D termites simulation but has no built in 3D visualization functionalities However its pure Java implementations enable Java 3D API integration 2 1 2D Termites Simulation 2D termites simulation is a common practice for self organized system research It is included in our software as a sample project We based our development on the simulation of collective building of 2D termites colony which involves two major objects termites and their woodchips 10 The termites gather wood chips into piles following a set of simple rules demonstrated in the StarLogo project 10 1 Each termite wal
15. t the colony is clearly viewed from different angles figure 6 7 Figure 7 Figure 6 We present a simple example of architectures grown by artificial agents moving randomly in the 3D space and performing simple asynchronous actions with purely local information 1 as shown in figure 8 Pr Figure 8 4 HARDWARE SPECIFICATIONS AND IMPLEMENTATION 4 1 Hardware Specifications In order to achieve the same results from Krieger s research 6 we needed to construct a small scale and low cost robot that can perform simple object removal tasks These tasks including moving on smooth surfaces detecting new objects woodchips in our case picking up an encountered new object and dropping the woodchip it carries when encountering another object The robot should come with sensors that are sensitive enough to detect objects within 20 30 cm range Because the final simulation requires a relatively large number of robot agents the robot should be easy and fast to assemble Since it moves around randomly it should be using batteries as its primary power supply a power cord will provide an extra obstacle For more economical reasons the robot toolkit should be reusable reprogrammable and consume as little power as possible It should be easy to connect to other devices We chose the Boe Bot Tool Kit from Parallax Inc and the Board of Education featured BASIC Stamp embedded microcontroller 18 4 2 Boe bot Descrip
16. tion The Boe Bot is built on a high quality brushed aluminum chassis that provides a sturdy platform for the servomotors and printed circuit board figure 9 18 Mounting holes and slots may be used to add custom robotic equipment 18 The rear wheel is a drilled polyethylene ball held in place with a cotter pin Wheels are machined to fit precisely on the servo spine and held in place with a small screw 18 In our case to simulate the termite s woodpile building process each Boe Bot needs a gripper The main controller of Boe Bot 1s a BS2 IC BASIC Stamp 2 which is a customized chip from Microchip PIC 16CS57C The BASIC Stamp 2 has 16 I O pins 2 dedicated serial port pins 1 input 1 output and room for 500 to 600 lines of code Detailed technical description for BS2 IC as well as its schematic can be found in Boe Bot User Manual 4 3 Termite Description In the following section we define the term termite as a Boe Bot with a pair of whiskers and gripper figure 9 The hardware design methodology divides the implementation into two major components whiskers module and gripper Module Left Whisker Right Whisker Ground o5 Fo Ground P6 P4 Figure 10 Whiskers schematic 4 3 1 Whiskers Module The Whiskers are used as object detectors since the BASIC Stamp can be programmed to detect when a whisker is pressed Once the termite touches a new object by it whiskers it will release the
Download Pdf Manuals
Related Search