coen-2003-project-24.. - Computer Engineering
Contents
1. return CView PreCreateWindow cs VEMM MM MOULE MMLlbbl bl lll EL lb P B M aM MM M P MM MAL t b b ll AAA P P AAA AAT CMappingView drawing void CMappingView OnDraw CDC pDC CMappingDoc pDoc GetDocument ASSERT_VALID pDoc draw VIL MMMMMELL BMC MM MB MUI MM Mall Vl P Ml lll AAT 11111 1 CMappingView diagnostics ifdef _DEBUG void CMappingView AssertValid const CView AssertValid void CMappingView Dump CDumpContext amp dc const CView Dump doc CMappingDoc CMappingView GetDocument non debug version is inline ASSERT m pDocument IsKindOf RUNTIME CLASS CMappingDoc return CMappingDoc m_pDocument endif DEBUG III SISI S 11S 111111111111111 111 1 CMappingView message handlers Thread consisting of the moving algorithm UINT Move LPVOID paran f bool left true bool turn false int i 0 PostMessage HWND param WM SCANDRAW 0 0 while true while red checkFront if left 33 red rotate 15 else red rotate 15 PostMessage HWND param WM SCANDRAW 0 0 turn true red move 10 PostMessage HWND param WM SCANDRAW 0 0 if turn if i gt 10 if rand 2 0 left true else left false turn false return 0 void CMappingView OnLButtonDb1C1lk UINT nFlags CPoint point HWND hWnd GetSafeHwnd if moveThread 02. pthread AfxBeginThread Move hWnd THREAD PRIORITY NORMAL moveThread 1 1 thread running else if moveThread 2 pIhread ResumeThread moveThread 1 1 thread running 34 else pIhread SuspendThread moveThread 2 2 thread suspended red stop CView OnLBut tonDb1Clk nFlags point void CMappingView OnRButtonDblClk UINT nFlags CPoint point scan draw for int x 41 red rot while scan draw x507 x ate 10 red turnDone CView OnRButtonDb1Clk nFlags point void CMappingView draw CClientDC dc dc FillRect C xpos red ge ypos red ge xpos conver ypos conver dc FillRect C for int x 0 this Rect xpos size ypos size xpost size ypos size white Xpos y EX EY EX ty ypos Rect xpos size ypos size xpostsize ypostsize black x lt xmax X 35 for int y 0 y lt ymax ytt if DB x y gt filter dc FillRect CRect x size y size xt size y size blue void CMappingView dataB for int x 0 x lt 8 amp amp sonars x 0 9999 ga sonars x 1 9999 x int pointx sonars x int pointy sonars x x int length sonars double probability pointx convertx pointx pointy converty pointy DB pointx pointy probability 1101 1 1l21 Gauss voi
3. I Comparative Tests of the Sonar data to actual distance of the robot to the wall Il Comparative tests of the Sonar data at different angles to a wall to the actual distance from the sonar sensor to the wall itself III Comparative tests in different locations in a room with e g corners noisy rooms and long halls Iv Next series of tests will repeat steps I III with the use of the filters Movement Mapping Our second step was to test out our movement algorithm once we had our sonar sensor data working correctly The progression of the testing attempted to start with a simple environment to a highly complex environment so that we can specify the limitations of the robot no matter what the case I Mapping of a square enclosed room with 4 walls IL Mapping of a square enclosed room with a movement algorithm that turns left only III Mapping of a square enclosed room with a movement algorithm which turns randomly IV Mapping of a square enclosed room with wander function with an obstacle 17 V Mapping of a hallway using the wander function starting from parallel to the walls VI Mapping of a hallway using the wander function starting from a random angle to the wall 4 2 Test Results Sonar Extensive testing has concluded a few parameters that our design should try to follow First of all our best readings came from when the sonar sensor is perpendicular to the wall it was pinging off of Slight change in the angle
4. As we analyzed earlier we received the best readings whenever the sonar is perpendicular to an object Dynamic thresholding would be another aspect of future design as we hope the robot would be able to determine the best degree of filtering by assessing the type of room it is in then optimizing the filter to get the best fit map Localization would also be ideally implemented by fusing the odometer sensor along with the sonar to compare and contrast 27 the previous taken values with the current values to make a final determination that the point it is at actually is the point in reality 7 3 The Big Picture Overall we would like to see this project continued and improved with years to come Hopefully even combine with other projects such as the formation robots Further development would allow further research into the optimization of sonar technology and perhaps a better way to implement it in industry We were fortunate enough to be able have the accessibility to these great learning tools the amigobots and hope others in the future will take full advantage of these robots in years to come 28 Refrences 1 Autonomous Robot Vehicles I J Cox and G T Wilfong eds Springer Verlag 1990 Collection of papers on various issues in mobile robots 2 Mobile Robot Self Localization without Explicit Landmarks R G Brown and B R Donald http www cs dartmout edu brd papers html Russel 3 Mobile Robot Localizati
5. Differential Gear ratio 19 5 1 Swing Radius 33cm Turn Radius O cm Translate speed max 750mm sec Rotational speed max 300 degrees sec Traversable step max 1 5 cm Traversable terrain all wheelchair accessible e Sensors O Fixed sonar 15 O 8 total e 4 forward 2 at rear 1 at each side Firing rate 25Hz 40 ms per sonar Sensitivity 10cm 3meters 6 inches 10 feet e Note any objects closer than 10 cm are read as 10 cm Problems include external noise and false echoes Limitations environment must be quiet flat surfaced with minimal reflected surfaces surrounding the perimeter Position encoders 2 one at each motor e 9 550 ticker per wheel revolution e 30 ticks per mm e Electronics O O Processor 20 MHz Hitachi H8 Position inputs 4 Sonar inputs 1x8 multiplexed Digital I O 8 bit logic ports A D 4 0 5VDC Digital timer inputs 6 microsec resolution Comm port 3 RS 232 serial FLASH 64 KB uP 1MB external RAM 16KB uP e Controls and Ports Main Power Robot accessories power ON OFF Charge System power battery recharge Reset Warm reboot download Mortors Test Motors download self tests Radio Power and serial Speaker 8 ohm Serial comm Ports 3 x RS232 Control System AUX 16 CHAPTER 4 TESTING 4 1 Test plan Sonar Our first step in testing the robot is to test out its sonar systems and the accuracy of them For this we conducted 2 different tests
6. returns true if the robot is currently connected endif defined AFX MYROBOT H CB03EFEO 5FOF 11D7 9B52 00E0988A2243 INCLUDED
7. in one direction We remedied this problem by implementing a random turning function where every ten iterations the robot would decide randomly to turn left or right This gave us map in which ensures that all parts of the physical world are probed Our wander function is somewhat similar to the simulation wander function found in the Activemedia suite We analyzed that the robot would only move when one of the front sonars and side sonars were reading open space in front of it Tuning the velocity of the robot turns and movements was also another challenge Fast abrupt movements would cause the robot to slip and corrupt more of our readings Secondly the corner distribution was unavoidably bad in most cases with filtering It is also important to note that the high filtered values cleaned up the very inaccurate readings from our sonar much better than the low filtered did Echoes that bounced around 2 3 different walls back to our sonar usually gave us readings further than what the real world was these values however were filtered out and the corners were cleaned up as you can see in the following pictures that display the same room with different grades of filtering Thirdly our mapping function in conjunction with our obstacle avoidance algorithm showed promising results as shown below DEG JEJ aot ET wv f mw UES i Si r 4 1 19 As you can see the environment with an oblong object which the sonar has a very hard time
8. slight angles We counteract these inaccuracies with the use of a Gaussian filter and distance filters We identified that the sonar works best at a certain threshold usually ranging from 1200 1500 centimeters which we have reflected to our filter This threshold also changes when the robot is exposed to different environments Consequently to make sonar more reliable dynamic thresholds in combination with a wall following algorithm will keep our values as accurate as possible and give us a better view of the real world Acknowledgements Special Thanks to Professor Neil Quinn and Professor Chris Kitts For their guidance and support on this project TABLE OF CONTENTS Page PADS IAC Geox rout kanda taya ts vases cae Lapang pba RP Hi Chapter d Introduction sess KNANG vete NAA NA NGA ANNE PAANAN NN ER NANANG NGA eU DERE TE 7 IT Backoround serine ea rep nate a d Metros 7 1 2 Project POalS M 8 1 3 Cntr DUU ONS cocco ene exeo C qox veru KAAGAD 8 Chapter 2 System Overview eve sadice qute ster Idus AB y quee EAEE EEE EEO VENDU ES 9 2 1 Use Be oce a ee OT e RO C dd e o eases 10 22 Conceptual MOdel ee iev ee euo ud coru ero PALANG Sou S e vvv EVER eeu vea KANO 10 2 3 List of Requirements functional and non functional esee 10 Chapter 3 Solbwales 2225 sees te REPE x x ede Tae ka EAE P ERA EDO E EE S PAANAN 11 Sl Architectural Diagrams 2 iu er cast cee ed nas soe Ne NANANA KANG a eee Pede 11 5 2 Class DOSCEIDU
9. to filter false sonar readings to our database Robot Diagram 1 1 Activemedia Amigobots Antenna signal amplifier as e 8 fixed sonars firing 25Hz 40ms per sonar e Sensitivity 10cm 3m 6 inches 10 feet oe Position encoders 2 one at each motor 9 550 ticker per wheel x revolution 30 ticks per mm 1 2 Project Goals Design Our main goal here is to do enough research to be able to develop a simple yet robust algorithm We attempt to take software engineers approach to this problem with little or no knowledge of sonar filtration or localization algorithms we find ourselves jumping into unknown territory It is imperative that our design be as flawless as possible or else we will be paying dearly in our testing phase trying to work out the bugs Implementation Activemedia has provided an extensive library of access functions to the robot known as Aria which is written in C We decided to write the rest of the mapping algorithm GUI and database all in C as well to ensure compatibility Another important concern for us during the implementation phase is whether will have sufficient time to debug the whole robot to an acceptable range of precision It s not known to us yet at how accurate the sonar will and how much error we will attain from the inaccurate nature of sonar technology Testing Different environments may give us very different results i e using sonar in a room full of loud subwoofers and aco
10. var return 500 A exp B 38 void CMappingView OnScanDraw WPARAM wParam LPARAM lParam scan draw void CMappingView OnRobotConnect if red isConnected red connect void CMappingView OnRobotDisconnect if red isConnected red disconnect void CMappingView reset for int z 0 z lt xmax zt t for int b 0 b lt ymax btt DB z b 0 0 clearScreen void CMappingView clearScreen CClientDC dc this dc FillRect CRect O O xmax ymax white Appendix A 2 Mapping View h MappingView h interface of the CMappingView class ISIS III II P II gg include myRobot h if defined AFX MAPPINGVIEW H 92E42C4C 5421 11D7 9B52 00E0988A2243 INCLUDED define AFX MAPPINGVIEW H 92E42C4C 5421 11D7 9B52 00E0988A2243 INCLUDED if MSC VER gt 1000 pragma once endif MSC VER gt 1000 const WM SCANDRAW WM USER 100 class CMappingView public CView protected create from serialization only CMappingView DECLARE DYNCREATE CMappingView double DB int xmax ymax xstart ystart size xpos ypos box double sonars double mean stdev PI var filter CWinThread pThread int moveThread CBrush black blue white Attributes public CMappingDoc GetDocument Operations public Overrides ClassWiza
11. IN MMPP P ge M PPP PPP TATA TATA MINE PE P P P P CMM I P P P TAA ATTA TATA ATTA TTT TTT fif defined AFX MYROBOT H CB03EFEO 5FOF 11D7 9B52 00E0988A2243 INCLUD pd P define AFX MYROBOT H CBOGEFEO 5FOF 11D7 9B52 00E0988A2243 INCLUDED dif MSC VER gt 1000 fpragma once endif MSC VER gt 1000 class public theta myRobot myRobot constructor virtual myRobot deconstructor void move int distance moves the robot forward if distance gt 0 and backwards if distance lt 0 void rotate int theta rotates the robot counter clockwise for theta gt 0 and clockwise for lt 0 double scan void using the center of the robot as the relative position ceturns and array of x y coordinates of the detected object and the distance the sonar was from the object int connect void connects to the robot int disconnect void disconnects from the robot bool turnDone void returns true if the robot is done turning bool moveDone void returns true if the robot is done moving void stop void Stops the robot from moving and turning 49 bool checkFront void returns true if there is something infront of the robot int getX void returns the X position of the robot where the front of the robot is north int getY void returns the y position of the robot where the front of the robot is north bool isConnected void
12. ODS dcs ouo et erre bu tek e bo pet epp RE REP ANNA ALA AG 11 2 9 Techiplgbie s USt AP POE TEENE 14 J4 RODOE Specifications S ence KANAN NPETE EEEE NG NANA EU xU EAMUS 15 Chapter d Testing AAP E OR EE REEE R Rade 17 4 1 Test Plan including test CASES sai dos re rre GNG GAGA 17 4 2 Test and Experimental Results eve ecu NA AGANG GAANO 18 Chapter 5 User ManU al icio se iige trot ue itat HR RENTES TR ven ER HA PRU a ER 22 Chapter 6 Other 5SUBSC disset terr erit eb x AHE NN toto aber bebe yeso dett er seems 24 Chapter 7 Conclusions a caves ye ko tea Saves e Ra SAY Y Ye VERS Caley YR SATA VYASE Ove aes 26 EN nra WORK TEE I LII I E T 26 7 2 Project Challenges Future implementation cesse 26 TI The DIS picture Y o desee oe er he ox dene BA pH E 27 NG AA AA eee 28 Appendix A Mapping Robot Code csse ehe eene hene 29 Appendix A l Myrobot epp ueeecevee tete eo rne nen ren Exe PEN GA AGA 29 Appendix A 2 Myrobot h 2 GANG NEGA GNG NGIGING UNA MEPR 36 Appendix A2 Mapping View epp aces eee aea aet nera Vet HERE LAN VIENNE EEOIT ve 38 Appendix A 2 Mapping View bi 22 kang dL eee LANGKA ae vx EUR ERU verd e dub ues 42 List of FIGURES Page Figure 1 1 Activemedia Amigobot Diagram sess y Figure 2 1 Use ase DAAN eii eo rne sot norte ddeewebes piden opui ident every Figure 2 2 Conceptual Model Diagram o eoe sor reto emus erp Des uc etal deine m 10 Figure 3 1 Architectural Diagram urn a
13. Santa Clara University DEPARTMENT of COMPUTER ENGINEERING Date June 11 2003 I HEREBY RECOMMEND THAT THE THESIS PREPARED UNDER OUR SUPERVISION BY Benard Farrales and Francis Chan ENTITLED Mapping Robot Utilizing Amigobots BE ACCEPTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF BACHELOR OF SCIENCE IN COMPUTER ENGINEERING Dr Neil Quinn Jr Christopher Kitts Dr Daniel Lewis Mapping Robot Utilizing Amigobots by Bernard Farrales and Francis Chan SENIOR DESIGN PROJECT REPORT Submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Engineering School of Engineering Santa Clara University Santa Clara California June 11 2003 Mapping Robot with Activemedia s Amigobot Bernard Farrales and Francis Chan Department of Computer Engineering Santa Clara University 2003 ABSTRACT Robots in today s world are utilized in environments that are just too dangerous for human intervention They provide us eyes ears and touch to places we usually see unfit Our objective is to provide a cognitive sense of mapping memory to assist in the explorative endeavors of these environments with the use of sonar technology In this paper we utilize sonar technology to map a room of unknown dimensions along with autonomous navigational capabilities Our experiments show that sonar data can be very inaccurate at very close and far ranges along with
14. after we readjusted the turning algorithm The filter is works well but needs to be readjusted depending on what kind of environment the robot is mapping The user will be able to choose a best fit map by turning the filter up or down The visual mapping part of our project works well and maps the data values stored in our database straight to the user with no problems 7 2 Project Challenges and Future Implementation Some of the challenges of our project included the limitations of sonar technology as it doesn t operate very well in certain environments especially that when the object to be read is too far too close or too high of an angle Obstacle avoidance was a challenge to find the exact turn radius interval to allow the robot to not hit a blind spot found within the sonar Whenever the robot did make contact with a wall we usually experienced wheel slippage as well which corrupted our sonar values It is also important to note that the velocity of forward movement and turning should be reasonably low as high velocities would cause the robot to jerk when it moved causing more wheel slippage in the process Finding the correct threshold filter for our test environment was challenging as well Design aspects allow us to change this to suit what ever environment we may run into decided upon the user Future implementation would include the ability to keep the sonar perpendicular to the wall to provide us the best values possible from our sonar
15. am works allowing the user to make new changes to the design if needed 8 Environmental impact Our product does not have an direct impact on the environment although it important to note that most of the materials on our robot except the power source are recyclable 9 Usability As noted the design of the robot interface was for simplicity and ease of use Basically all the user needs to do is push one button after uploading the code 10 Lifelong Learning The design on this robot could be developed in many other ways allowing those who will take on this project in the future to learn even more about robotics and the way we replicate our human senses to the artificial world Multiple threading with other robots and further development on ways to filter out sonar data are just a few more things we could learn more about with further implementation 11 Compassion Our project if developed further would be able to help the blind navigate through environments without the use of a walking stick or walking dog Our robot could actually serve as a type of seeing eye dog that could be able to translate the maps into brail as well 26 CHAPTER 7 CONCLUSION 7 1 Summary of Work After further testing the flaws of our code were worked out so that our final working product is able to move filter and map The movement algorithm proved to be very reliable as our incident of impact to any obstacles in a room was minimized to zero
16. ase and generates a GUI of a map Robot Command control for our robot which receive and deliver commands sent via wireless modem Movement Our movement algorithm consists of a wander function as described in the following state chart 3 4 Nothing in front Something in front Rotate right Move Random forward Function Rotate left 13 3 3 Technologies Used Aria o Library functions for the amigobot provided by activemedia that allows us to communicate with the robot Activemedia software suite o Provides us interface to programming and observing the robot Rhapsody o C UML program used to construct our OMDs and statechart diagrams Microsoft Visual C o Used to develop our code and run our robot Amigobots by Activemedia Robotics o Robotic unit o Specifications included on next page 14 3 4 Robot Specifications Mapping Amigobot SPEC e Physical Characterisitics O O O O Robot measurements 33cm X 28cm X 13 cm Body clearance 3 cm Weight 3 6 Kg Payload 1 Kg e Construction O O O e Power O O O O O Body Molded polycrb Chassis 1 6mm fab alum Assembly Allen hex screws metric 12V lead acid Charge 24 2 watt hr Run time 3 hours Recharge time trickle 12 hrs Recharge time hi cap 4 hours e Mobility O O O O O Drive wheels 2 solid rubber with caster balance Wheel diameter 10 cm Wheel width 3cm Steering
17. d CMappingView Decrement int x int y DB x y1 710 void CMappingView scan int xpos2 red getX int ypos2 red getY xpos2 convertx xpos ypos2 converty ypos Decrement xpos2 ypos2 sonars red scan dataB LA LA LA i void CMappingView OnKeyDown UINT nChar switch nChar reset DB case 46 delete reset break scan and case 9 scan draw draw 0 0 break draw spacebar case 32 draw brea scan en case 13 scan brea move forward case 38 0 K ter 0 Ky tab up key red move 10 whil brea move bac case 40 K kwards red moveDon down key red move 10 while red moveDon break rotate left left key case 37 red rotate 10 while red turnDone break 0 1 0 1 UINT nRepCnt UINT nFlags 37 cotate right right key case 39 red rotate 10 while red turnDone break CView OnKeyDown nChar nRepCnt nFlags void CMappingView OnKeyUp UINT nChar UINT nRepCnt UINT nFlags red stop CView OnKeyUp nChar nRepCnt nFlags int CMappingView convertx int i i box i xstart return i int CMappingView converty int i i box i ystart return i double CMappingView Gauss int x double A 1 stdev sqrt 2 PI double B x mean 4 x mean 2
18. e values somewhat Our results however turned out fairly decent as both the doorway and the walls were mapping with decent precision Our last test was conducted in a random environment where no artificial biases could be presented in favor of the sonar The results are as follows 4 4 4 5 rm PI EI I Eu NS KN ware FEJA nuw With medium filtering the results we retrieved were very promising Because of large number of oblong objects and loose material i e electrical wire found on the ground we were expecting very poor values But in fact the robot was able to scan the garbage can found on the left which is the round half moon values taken on the left side of the wall 21 Many of the loose wiring found at the end of the room was scanned fairly accurately and the robot had no problems navigating through this environment 22 Step 1 Step 2 Step 3 Step 4 Step 5 Step 6 Step 1 Step 2 Step 3 Step 4 CHAPTER 5 USER MANUAL SOFTWARE INSTALLATION 1 Install Aria Active Media Suite and Visual C on your computer In Visual C create a new MFC Application project titled Mapping using the Wizard and the default settings Link the new project to the Aria libraries See Sample Aria projects for information on how to link libraries Download myRobot cpp and myRobot h and import them into the project workspace Replace Mapping View cpp and Mapping View h with the updated ve
19. etVel vel robot move distance 43 while robot isMoveDone void myRobot rotate int theta robot setRotVel rotVel robot setDeltaHeading theta while robot isHeadingDone double myRobot scan void int z cursor 0 int y robot getNumSonar double sonars new double y for z 0 z lt y z f sonars z new double 3 for z 0 z lt y z for int b 0 b lt 3 btt sonars z b 9999 for z 0 z lt y z f retrieves the position and angle of a specified sonar double distance theta int x y switch z case 0 x 105 y 73 theta 3 14159 break case 1 x 78 y 130 theta 2 33874 break case 2 x 30 y 154 theta 1 78024 break case 3 x 30 y 154 theta 1 36136 break case 4 x 78 y 130 theta 0 80285 break case 5 x 105 y 73 theta 0 00 break case 6 x 60 y 146 theta 5 340708 break case 7 x 60 y 146 theta 4 084070 break distance robot getSonarRange z adds sonar data only if the distance is between if distance gt 300 amp amp distance lt 1500 300 and 1500 45 theta theta robot getTh 3 14159 180 x 1 robot getY x distance cos theta y robot getX y distance sin theta sonars cursor 0 x sonars cursor 1 y sonars cursor 2 distance cursortt return sonars int myRobot getX return 1 robot getY int myR
20. ngView CMappingView axis axis red connect filter 0 0 minimum probability in order for coordinate to be valid drawn mean 800 mean for the gaussian function stdev 200 standard deviation for the gaussian function PI 3 14159265 var stdev stdev xmax 1020 width of the screen in pixels ymax 675 height of the screen in pixels size 2 size of the dot to be drawn at coordinate x y box 12 increase the number to lower the resolution xstart xmax 2 x starting position of the robot on the screen in this case the center of the x ystart ymax 2 y starting position of the robot on the screen in this case the center of the y xpos xstart ypos ystart moveThread 0 initialize at 0 to indicate no thread is running creates 3 brushes black blue and white black new CBrush 31 black gt CreateSolidBrush RGB 0 0 0 blue new CBrush blue gt CreateSolidBrush RGB 0 0 255 white new CBrush white gt CreateSolidBrush RGB 255 255 255 creates a two dimensional array of xmax and ymax and initializes the values to 0 DB new double xmax for int z 0 z lt xmax ztt DB z new double ymax for z 0 z xmax zt t for int b 0 b lt ymax btt DB z b 0 0 CMappingView CMappingView if red isConnected red disconnect BOOL CMappingView PreCreateWindow CREATESTRUCT amp cs
21. obot getY return robot getX int myRobot connect Aria init Aria setKeyHandler amp keyHandler robot attachKeyHandler amp keyHandler serConn setPort robot setDeviceConnection amp serConn robot addRangeDevice amp sonar if robot blockingConnect printf Could not connect to robot exiting n 46 Aria shutdown return 1 turn on the motors turn off amigobot sounds robot comInt ArCommands ENABLE 1 robot comInt ArCommands SOUNDTOG 0 robot comInt ArCommands SONAR 1 start the robot running true so that if we lose connection the run stops robot runAsync true return 0 bool myRobot turnDone void return robot isHeadingDone bool myRobot moveDone void return robot isMoveDone void myRobot stop void n robot setVel 0 robot setRotVel 0 bool myRobot checkFront void int range for int x 0 x lt 5 x t range robot getSonarRange x if range lt minDistance return true return false int myRobot disconnect void robot setVel 0 Aria shutdown robot disconnect return 0 bool myRobot isConnected void return robot isConnected Stop the robot Shutdown Aria Disconnect from the robot returns true if the robot is connected 48 Appendix A 4 MyRobot h myRobot h interface for the myRobot class MILL M
22. obot MCU e Closed controlled static environment to be mapped e C compiler 10 CHAPTER 3 SOFTWARE 3 1 Architectural Diagram 3 1 movement 3 2 Class Descriptions Database consists of a 2 dimensional array of x y coordinates Each value is tagged along with a probability value determined by our filter Filter The filter is based upon two functions the first one being a simple distance filter given to us by the limitations by the sonar Specifically the sonar will not register if it is too close or too far from an object and we have accommodated this discrepancy when ever we take readings with a threshold The second filter we have implemented is a Gaussian function 11 f z Le mre i a 2m l After extensive testing of the robots sonar we pinpointed a value in which we accumulated the best readings and used that as our peak Hence whenever we took a value the closer to our peak a higher probability value was tagged onto the sonar reading to be stored into the database The end result is a valley of sonar readings to be stored into database as shown in the next diagram 0 6 04 3 3 12 Hence the end result will be a map consisting of valleys in which we can decide exactly how much of the valley we want to show on the map which will translates to a higher filtered map the higher probability we choose to use vice versa Map takes in the filtered values from the datab
23. on and Map Building A Multisensor Fusion Approach J Catellanos and J Tardos Boston MA Kluwer Academic Publishers 2000 4 Wall Following Using Angle Information Measured by a Single Ultrasonic Transducer Teruko Yata Lindsay Kleeman Shin ichi Yuta http www ri cmu edu pub files pub3 yata teruko 1998 l yata teruko 1998 1 pdf 29 Appendix Appendix A l Mapping View cpp MappingView cpp implementation of the CMappingView class include stdafx h include Mapping h include MappingDoc h include MappingView h include lt math h gt include lt process h gt include lt windows h gt ifdef _DEBUG define new DEBUG_NEW undef THIS FILE static char THIS FILE __FIL endif Gl myRobot red III 111111111111111 111111111 111111111111111 CMappingView IMPLEMENT DYNCREATE CMappingView CView BEGIN ESSAGE MAP CMappingView CView AFX MSG MAP CMappingView N WM LBUTTONDBLCLK RBUTTONDBLCLK KEYDOWN KEYUP Z Z Z Z QOOOo0onR Zz S ON_MESSAGE WM_SCANDRAW OnScanDraw ON_COMMAND ID_ROBOT_CONNECT OnRobotConnect ON_COMMAND ID_ROBOT_DISCONNECT OnRobotDisconnect AFX MSG MAP END M ESSAGE MAP VALLALLA AALA LALALALA ALALLA TATA TATA AAT AAA TTA AAA ATTA ATTA TATA ATTA TTT TTT CMappingView construction destruction CMappi
24. pCnt UINT nFlags afx msg void OnRobotConnect afx msg void OnRobotDisconnect afx msg void OnScanDraw WPARAM wParam LPARAM lParam user created message to execut scan and draw during a thread AFX MSG DECLARE MESSAGE MAP bi ifndef _DEBUG debug version in MappingView cpp inline CMappingDoc CMappingView GetDocument return CMappingDoc m_pDocument endif 111111111111 AFX INSERT LOCATION Microsoft Visual C will insert additional declarations immediately before the previous line endif defined AFX MAPPINGVIEW H 92EF42C4C 5421 11D7 9B52 00E0988A2243 INCLUDED 42 Appendix A 3 MyRobot cpp myRobot cpp implementation of the myRobot class SISI III PIE ME MM MP P TATA TTT TTA TTA TTT nclude stdafx h nclude Mapping h nclude myRobot h nclude Aria h nclude lt fstream h gt H H p H H ifdef _DEBUG undef THIS_FILE static char THIS_FILE FILE define new DEBUG NEW endif III 111111111111111 III III 1111 11 Construction Destruction 111 111111111111111111111111111111111111111111111111111111111111111 ArRobot robot ArSerialConnection serConn ArSonarDevice sonar ArKeyHandler keyHandler double vel 30 double rotVel 10 int minDistance 350 myRobot myRobot myRobot myRobot void myRobot move int distance robot s
25. rd generated virtual function overrides AFX_VIRTUAL CMappingView public virtual void OnDraw CDC pDC overridden to draw this view virtual BOOL PreCreateWindow CREATESTRUCT amp cs protected AFX_VIRTUAL Implementation public virtual CMappingView ifdef _DEBUG virtual void AssertValid const virtual void Dump CDumpContext dc endif protected void scan position void dataB void draw int convertx int 1 int converty int i double Gauss int x void Decrement int x int y void reset void clearScreen Generated message map functions protected AFX_MSG CMappingView const retrieves the position of the robot decrements that in the database and passes the sonar data to the database increments the coordinates of the database depending on the returned value from the gaussian function draws points depending on the values in the database converts the x value to be stored converts the y value to be stored returns a value between 0 and 1 depending on the values x mean and stdev decrements the value of the coordinates x y sets all values in the database to 0 clears the screen afx msg void OnLButtonDb1C1k UINT nFlags CPoint point afx msg void OnRButtonDblClk UINT nFlags CPoint point afx msg void OnKeyDown UINT nChar UINT nRepCnt UINT nFlags 41 afx msg void OnKeyUp UINT nChar UINT nRe
26. reading because of its round shape is mapped fairly well considering its structure The four walls are defined in this map using medium filtering Another important aspect to this filtering is that it can tuned in many ways The sensitivity of the sonar itself is adjustable so that better readings of long flat objects can be taken accurately while smaller objects are more likely missed The opposite is true when the sensitivity is turned up Once the map has been attained the filter can be turned up or down in order to optimize the map of the real world Our long hallway mapping test went very well whenever we Started our robot at the ideal perpendicular position as shown below This test was taking over a 5 minutes time span down a long hall found in the engineering dept The walls were constructed of drywall and as you can observe the walls are very accurately mapped along with the doorways However this won t be the case all the time when the robot is placed in a random environment so also duplicated this experiment with the robot at a biased angle with a medium fitler 4 2 However this won t be the case all the time when the robot is placed in a random environment so also duplicated this experiment with the robot at a biased angle with a medium filter 20 i amp 7 Nag is PETE BAE SSS EE Sees VSR dum L WET A E n Tamod Po RT Hew 4 3 As you can see the smallest variation in angle can throw off th
27. rk Ee RR ER ENSE E Ex Y MERE TREE Hats 11 Fig re 3 2 Gaussian Pune ion 229 etek acs ck t NG ADARNA NINANG dea 12 Figure 3 3 Database SIMULALON iios pres og Dn GANANG NPA GAD MANA Sabes ABANGAN 12 Figure 3 4 Movement Statechart 1i panda NG NG eps ex Xx AALAGA 13 Figure 4 1 Test case with Oblong Object in middle of control environment 19 Figure 4 2 Test case down long hallway c cocuse cence ence kana me 20 Figure 4 3 Test case down long hallway at different angles usa saaan asan 20 Figure 4 4 Test case in random rOOM 6 eee scence cece ee eee ence I eI mene 21 Figure 4 5 Picture of Random room cca ccs ore eee YNNA NA 21 Chapter 1 INTRODUCTION Our main goal of this project is to allow a robot equipped with sonar sensors to map a room with unknown dimensions at an unknown origin 1 1 Background The primary function of our robot is for us to be able to drop our robot off anywhere in an unknown room and allow it to navigate throughout the room while incrementally mapping its pinged sonar values to a 2D plane The bundled Activemedia software already provides us with a wide array of functions for our robot however it lacks the robust autonomy to actually map an unknown area Main Objectives e Development of an interface between our mapping software and the robot e Development of an GUI interface and graphical mapping program e Development of an algorithm that will be able
28. rsion Compile and Run the code QUICK INSTALLATION Create a new Folder and copy the Mapping Directory into it Install Aria Active Media Suite and copy the libraries into the newly created folder Open up the Mapping Workspace using Visual C Compile and Run the code 23 CONNECTING AND DISCONNECTING FROM THE ROBOT You should automatically be connected to the robot upon execution of the program If by chance you are not click on the Robot menu and select Connect To disconnect from the robot simply close the program or select Disconnect from the Robot menu CONTROLLING THE ROBOT up arrow move forward down arrow move back left arrow rotate counter clockwise 10 degrees right arrow rotate clockwise 10 degrees enter Scan sends the sonar data to the database spacebar draws a map using values from the database delete clears the data in the database and clears the screen AUTOMATED FUNCTIONS Wander double left click Wanders around the room while scanning and avoiding nearby objects and continually updating the map Double left click again to pause and resume Spin and Scan double right click Rotates 360 degrees counter clockwise while taking scans every 10 degrees 24 CHAPTER 6 Other Issues 1 Social Social impact of our project would allow society to venture into hostile environments which would be normally accessible for human interaction For example unsafe building
29. s could be probed by our robot for safe paths to travel Hazmat teams could be able to search buildings for obstacles and safe paths to traverse in order to find precious cargo Essentially it will allow society an artificial cognitive sense of mapping as robots have allowed us to extend our senses without putting ourselves in immediate harms way 2 Political Our project has no political aspect what so ever 3 Economic Our project replicated a costly piece of software that is sold by Activemedia and could be easily further developed to be competitively held up against Activemedia s mapping software In a sense we are on the road to stimulate the market for robotics software 4 Health Our project has not direct relation to any current health issues 5 Safety Our project is relatively safe The mapping software allows people to map environments and get a better idea what their physical world looks like without them getting out of their chair 6 Manufacturability The Mapping robot software can be easily ported to any amigobot making it very accessible to anyone who owns both an amigobot and the activemedia software bundle 25 Manufacturability is only limited to how many robots Activemedia produces per year and different prices 7 Sustainability We developed the software to be very easily used and maintained Documented comments and simple user interface will give the user a better conceptual idea of how our progr
30. s however would throw our readings off especially at very close and far ranges Corners also served as a problem for the sonar as readings that came from them usually read either very close or very far beyond the actual physical object One thing we did notice is that readings from a 120cm 150cm on average seemed to give us the best readings We used values within our threshold Gaussian function as the peak point The closer the sonar readings were to these distances the more we can believe that they were actually there Of course our values in which were below 50cm and beyond 2000cm were automatically thrown out of our database even though they were read in Further results of our filter in action will be more extensively shown below Different materials from hard wood walls to dampened dry wall seemed to give us both the same type of readings consistently Even the poor readings we got in other test cases were received on different mediums For corners we realize that it just simply the downfall of sonar technology and that certain levels of inaccuracy can be expected Movement and Mapping Initial results with our moving algorithm started with a simple left turn only algorithm We quickly changed this when we noticed that the robot would keep moving in the same circle around our closed control environment The results of the continuous circles would 18 cause our map to drift counter clockwise because of the continuous wheel slippage
31. ustics will render our readings useless We will have to provide enough diversity in our testing phase to ensure that all the algorithms are working efficiently and give us accurate map values Finding an area where we will able to conduct these experiments accurately 1 3 Contributions We would like to thank the Santa Clara University Robotics department for the donation of the amigobot pc s and lab rooms to make this project possible CHAPTER 2 SYSTEM OVERVIEW 2 1 Use Case 2 1 robot_control activates n runs Database User E The Robot Control will hold our mapping and movement algorithms that we programmed into the robot all the user has to do is activate and run the robot The user will interact with is the database which will contain a flat file of x and y values The display map will have a visible GUI interface that will graphically display the values in the database in a visual topological grid User Activates and runs the robot User has the option to analyze raw data values in database User can visually analyze the map via gui 2 2 Conceptual Model 2 2 MovingAlgorithm Commands SonarVals Polar to Rectangular transformation X Y coordinates amp Distance from objects FilteredVals Database 2 3 List of Requirements e Sonar Odometer Equipped robot provided by Activemedia s Amigobot e PC w 300Mhz or better w open serial port e Software Interface to r
Download Pdf Manuals
Related Search