AI-robotprojektreport
Contents
1. Vad som ska g ras public void actionPerformed ActionEvent e I Object object e getSource if object quitButton kid turnLeft kid map getAngle kid emergencyBrake System exit 0 else if object collisionAvoidButton CA j else if object followLightOrWall followLightOrWall else if object mapNPos mapAndPos 15 else if object localize localize else System out println ERROR VLLEZLZLLLLLLLLLLLLLLEALLLLLLLLLLLLLLLLLE s END SJ LVA KODEN f AOI RO RU de ORCI OUR ROGER d class Table extends Canvas private Graphics graphics private int dx dy private int x y private int height width public int map public Table dx 4 dy 4 x 0 y 0 height 100 width 100 public void setCoords int x int y this x x this y yi public void setMap int map this map map public void update Graphics g g setColor Color white for int i height 1 i gt 0 i for int j 0 j lt width j if map j i 0 I setCoords j i g fillRect x dx 400 y dy ax dy class Robot private static final int LF 10 LineFeed avslutar alla Khepera reply private static final int DIA 55 Khepera s diameter i mm private static final double MOVEUNIT 0 08 Enhet som Khepera s hjul anv nder private static final double OMKRETS Math PI DIA priv2. for int i height 1 i 0 i System out print i for int je0 j width js if map j i 1 I ie System out print else if map j i 0 System out print else if j x amp amp i y System our print X System out println ri 28 public void setMapMaking boolean b makeMap b j public int getMap return map H public int 1 getModifiedMap int i j k 17 int modifiedMap new int width height for i 0 i lt width i for j 0 j lt height j modifiedMap il j map il jl for isl i width 1 i 4 for j l j lt height 1 j if map i j 1 for k 1 k lt 1 k for 1 1 1 lt 1 1 if k 0 amp amp 1 0 if map itk j 1 0 modifiedMap i j 0 return modifiedMap 29
3. isCorner true rakt fram else rightTurn false else if sens SENS L45 5600 amp amp sens SENS_L90 gt 850 leftTurn false Map making if makeMap map set TollowWall sens kollar om man n tt m let vid h rn if isCorner isCorner false if map reachedPos x y emergencyBrake foundPos true 18 for int i 07i lt nufCorners i System out print cornerMatrix i matrix CORNER System out printin originalCorners corner Matrix2String cornerMatrix System out println originalCorners toriginalCorners saveContext F rs ker lokalisare sig genom att st lla sig vid starth rnet vid mapmaking Detta g rs genom att titta p alla h rn och sedan j mf ra dessa h rn med h rnen man hade d man gjorde kartan mapmaking public void localize int sens boolean foundPos false boolean leftTurn false rightTurn false String newCorners boolean isCorner false int x mapWidth 2 int y mapHeight 2 goToCorner reset System out printin START LOCALIZEBSBG GJ Map localMap new Map mapWidth mapHeight localMap setMapMaking true nufCorners 0 F lj v gg while foundPos sens getProxSensors localMap updatePos getPos if leftTurn if sens SENS L45 KONVEX LIMIT leftTurn true rightTurn false cornerMatrix nufCorners matrix CORNER KONVEX cornerMatrix nufC
4. total distance on right wheel 55 radius R on the big circle and diameter on the robot R2z Circumference of a circle with radius R Fig 2 Angle calculation We needed to analyse whether the map representation and appreciative calculations were sufficiently corresponding to reality and therefore conducted some validity tests We let the robot explore a map and try to stop at the starting position and came to the conclusion that it would find home safely if home was defined 4 cm which means 2 squares in our grid Here we encountered another problem P4 The image of the map would be mirrored in the representation We adjusted the value definitions so that it would correctly display the map as it appeared in reality Homing behaviour Since the robot was also supposed to be able to find a position in the map and go to this point we needed a fairly good accuracy We made a complex formula for this kind of behaviour that would have to foresee that eventual objects were blocking the robot to move along a straight line from point A to point B anywhere anytime in a random labyrinth After some discussions we all agreed that the limited time forced us to postpone this perfect A to B behaviour and transform it into a simpler home position recognizing behaviour that our present robot would manage What strategy were we supposed to use We noticed that the other group were experiencing problems with trying to calculate average p
5. To use the lab time as effectively as possible through parallel programming and testing at the lab 3 To use a reactive model consisting of following modules Collision avoidance behaviour Wall following behaviour Map drawing behaviour Localisation behaviour Light source behaviour O oO 8 E Since the base task required for us to graphically visualize a map of the environment after the robot had explored the environment we discussed different methods to do so The robot needed to be able to explore an entire labyrinth no matter how it looked and also needed to be able to avoid objects in it s path Therefore we came up with a plan of heuristics that the robot could follow regardless of what kind of labyrinth it was put into Heuristic plan with add on modules Collision avoidance behaviour that enabled the robot to freely walk around and register data about the environment without being physically stopped by objects or walls Potential Draw a map randomly but may go into same areas over and over again 2 Follow wall behaviour which made the robot tune its sensors according to the walls and establish a smooth parallel movement alongside the wall it follows Potential This add on will make the robot explore all labyrinths without islands 3 Memorize and follow non visited objects When the robot notices other positions that clearly contains a object it memorizes it in a list so that it can go to this area and start
6. map Math round xCoord 250 origoX Math round yCoord 250 torigoY 0 Nord 1 x float xCoord k Math sin angle y float yCoord k Math cos angle map Math round x 250 torigox Math round y 250 origoY 0 NordOst 2 x float xCoord kDiag Math cos beta y float yCoord kDiag Math sin beta map Math round x 250 origox Math round y 250 origoY 0 OsETt33 7 x float xCoord k Math cos angle y f loat yCoord k Math sin angle map Math round x 250 origox Math round y 250 torigoY 0 Syd st 4 21 x float xCoord kDiagtMath sin beta y float yCoord kDiag Math cos beta map Math round x 250 origoxX Math round y 250 torigoY 0 Syd 5 x float xCoord k Math sin angle y float yCoord k Math cos angle map Math round x 250 origoX Math round y 250 torigoY 0 SydV st 6 x float xCoord kDiag Math cos beta ve float yCoord kDiag Math sin beta map Math round x 250 origoxX Math round y 250 torigoY 0 Vast 7 x float xCoord k Math cos angle y float yCoord k Math sin angle map Math round x 250 origox Math round y 250 origoY 0 Nordvast 8 x float xCoord kDiag Math sin beta y float yCoord kDiag Math cos beta map Math round x 250 origoX Math round y 250 torigoY 0 returnerar robotens nuvarande vinkel i grader j
7. ERE OR SEREEREREEEERERERE MERRIER AR KRRARARAR ER RAR KARA RAR RAR ARERR E EE BRK RRR EEE RR KIRKE RRR AA RR AKA RRR RRR KRKA L METHODS FOR CONTROLLING KHEPERA SZ LEE EE REE RRI KEKE ERE RIESE REE RINK ERERHEAAEKRER public boolean reachedPos int motionStat getMotionStatus return motionStat 0 T ON TARGET amp amp motionStat 3 T ON TARGET public void reset setSpeed 0 0 setPosition 0 0 public int getMotionStatus sendMessage K n return getintReply k 23 public void setSpeedlint leftWheel int rightWheel sendMessage D tleftWheel rightWheel n getReply speedLeft leftWheel speedRight rightWheel public void setPosition int leftWheel int rightWheel sendMessage G leftWheel rightWheel n getReply LY postionLeft leftWheel positionRight rightWheel public void setReachPosition int leftWheel int rightWheel sendMessage C leftWheel rightWheel n getReply postionLeft leftWheel positionRight rightWheel public void emergencyBrake sendMessage D 0 0 n getReply public int getProxSensors sendMessage NMn return getIntReply n I public int getPos sendMessage HNn return getintReply h public int getLightSensors sendMessage O n return getIntReply o public int getVisTurretSensors send
8. behaviour When the base goal was completed we had to go on with the main goal Which was to make the robot follow a light source For practical reasons we used a big flashlight and therefore could freely move it around in different sections of the labyrinth This behaviour was very similar to the Braitenberg method integrated in the wall following method We just added the light from the flashlight as another factor the robot had to adjust itself to When there was no light it would only use wall following behaviour and when the light appeared it would move towards it until the light was so strong that it would consider itself to have found the light source If the light then was abruptly removed it would just switch back to the wall following behaviour once again The only problem P7 we encountered in this section was that if the light happened to project onto the walls there may be a conflict between the lit up wall and some part of the lit up labyrinth ground The robot would then stand in some kind of deadlock and not know where to move If only the walls were lit up it would just continue straight on into the wall since it saw the wall as some kind of light it needs to follow We solved this by deciding that light should only come from straight above Constraint 5 After adding this constraint everything worked fine and the robot would smoothly move in the labyrinth until it would track some light source and move towards it 4 Problems soluti
9. f 500 S500 sens SENS R901 10 500 speedR cruiseSpeed int frontR sideR diagR if sens SENS L10 TOOCLOSE IW sens SENS L45 lt TOOCLOSE sens SENS_L90 lt TOOCLOSE II sens SENS RIO TOOCLOSE sens SENS_R45 lt TOOCLOSE sens SENS R90 TOOCLOSE foundLight true setSpeed speedL speedR 21 BERR RRA HERE ERK ERR ERR EHR IK EERE KAI RTR KARTA REED d RRR RARER RRR EERE KEK EER ERA KHER RA KARA KARI EER ERE RE ZS END BEHAVIORS EL BRR RAK FOR e e e ke ke e e e he hee e e ke hohe he e ee Ree e e dk RR e e BERR ERR IRR IKE RIK IK IRI KIER IS ER IPRA IK HHA EMR HEE RR RERKRRK AKER RC ACkCk kCACk RCk Ck Ch KCKCh KCKCKCkCK ER ERR KERR K REA EKE KEKE ER II I ROO eo ek ceo ie eke eck eid do kok de ke d Diverse hjalpmetoder S RRR IK RATA RR IA ARR RT IRR RITE SAR RRR IKK KK EERE HERA RR RRR RARE KERR RAR EAR k A h ACA OkOAOCACA Ch RARE RRERAKK KARR kk K kk k k kc d Sparar robotens stateinfo dvs dess hjulpositioner SCH public void saveContext saved robotPos getPos map saveContext public void restoreContext setPosition saved robotPos 0 saved robotPos 1 map restoreContext Analyserar matrisen med h rn och skapar en str ng med de h rn vi tror faktiskt finns Antar att sista och n st sista h rnet i matrisen ej r kopior public String corner Matrix2String int m int LIMIT 0 String retCorn
10. AI Robot Project Report DATI25 AI for Robots HT 00 Project 4 group Cl CL Bo Do Jeton Grajgevci Robert Hedlerfog Contents UK umen I8 r CL sn E ET FEE br SORAN ENAS TT 3 KN cain annus vissatanea rrit vik iv cada 3 d RARO en unis 1 zessan sins Meme gta a x S ait ae AR 3 3 Continuous working process of the robot project ckt 5 2 0 Rider OR BEBE EEN 3 3 2 Implementation phase c ENER ee eene 6 4 Problems solutions and constraints eee 10 S Results andthe final status of the robot oer ce ean ene GEI VIR EEN T Rubre WORK on tbe robot sses scesi soune ENNEN nena XV V ts E iov RDUM C MF 13 CRT 13 1 Introductory This is a report of our robot project work in course DAT125 Al for Robots HT 00 Our group C1 consisting of Bo Do Jeton Grajqevci and Robert Hedlerfog chose project no 4 with the following definition Hide and seek looking in a maze for a robot holding a light source on it We decided to extend the project if time would allow us The extension was to challenge another group and implement the ability to track another robot and then move home as quickly as possible 1 1 Method We decided to use the JAVA language since we all had some knowledge and interest in Java programming The communication protocol written by Robert Pallbo in Java was another argument for us to complete the project with JAVA We decided to us
11. CORNER KONVEX retCorn X else System out println Konstiga h rn Dubletter else E m nufCorners m nufCorners 22 if m 0 matrix CORNER KONKAV retCorn 4 O else if m 0 matrix CORNER KONVEX retCornt X else System out printin Konstiga h rn return retCorn returnerar hur m nga steg new ska cirkul rskiftas t v nster f r att se ut som orig public int compareCorners String orig String newCorn int difference 0 if orig length newCorn length System out printin OLIKA ANTAL H RN difference 1 System exit 1 else while orig compareTo newCorn 0 I newCorn newCorn substring 1 newCorn charAt 0 differencet return difference public int getMap return map getMap public int getModifiedMap return map getModifiedMap public void setMapMaking boolean b makeMap b map setMapMaking b Kollar om roboten fastnat i en terv ndsgr nd i s fall v nd 180 grader public void testZero if speedLeft 0 amp amp speedRight 0 nufZero if nufZero gt 5 emergencyBrake turnLeft 180 else nufZero 0 ARRARARARRARKAKAARARRARARRARA RR ee eee d e AU Ue e AR RAR RRR FIRAR RAR ARA RAA RR RR KRA RA RAA RA RAR AR AIR RAR KART RAR for END Diverse hj lpmetoder be BRI IRE IR AM TK RIK ERR RIE REE EERIE IT RIM RRA ERE CLE KH f ORO ode de ok deed e ERE
12. Message T 2 NMn return getIntReply t 2 n Robert Pallbo s Sends a message to the robot The reply is returned Failed transmission results in a warning printed on the consol public static void sendMessage String msg try outputStream write msg getBytes catch IOException e System out printin Warning Message could not be sent to robot System out printin Message was msg System out println e getMessage Robert Pallbo s Waits until data is available from the Khepera public static void waitForReply try while inputStream available 0 catch IOException e System out println Warning Waiting for reply failed System out printin e getMessage Robert Pallbo s Reads a reply or a message from the robot public static String getReply String reply Las in data fram till LineFeed try int b inputStream read while b LF 24 reply char b b inputStream read catch IOException e System out printin Warning ply from robot could not be read System out printin e getMessage Returnera resultatet return reply Robert Pallbo s Reads the reply from Khepera and puts the reply in an int array The prefix in the reply is ignored when parsing the reply public int getIntReply String prefix H L s in responsen som en str ng och plocka bort prefixet String repl
13. Sensors if leftTurn SF sens SENS L45 KONVEX LIMIT leftTurn true rightTurn false else if sens SENS L10 2KONKAV LIMIT if rightTurn rightTurn true corners t h else rightTurn false else if sens SENS_L45 gt 600 amp amp sens SENS_L90 gt 850 leftTurn false followWall sens emergencyBrake public void followWall int sens int speed speedL speedR float sideLeft diagLeft frontLeft y int sens boolean foundPos false FO lj v gg 4 7 sens getProxSensors frontLeft sens SENS L10 8f 1023f sideLeft sens SENS L90 500 2f 500 diagLeft sens SENS L45 800 5 500 minska 500 Ta bort speed int frontLeft sideLeft diagLeft Svang in t minus p v nster hjul if speed 0 speedL cruiseSpeed speed on speedR cruiseSpeed Sv ng ut t minus p h ger hjul else speedL cruiseSpeed speedR cruiseSpeed speed speed lt 0 setSpeed speedL speedR public void go2theLight int sens int speed speedL speedR float sideLeft diagLeft frontLeft float sideR diagR frontR int TOOCLOSE 50 frontLeft 500 sens SENS L10 2f 500f diagLeft 500 sens SENS L45 8f 500f sideLeft 500 sens SENS L90 10f 500f speedL cruiseSpeed int frontLeft sideLeft diagLeft Y og frontR 500 sens SENS R10 2f 500 diagR 500 sens SENS R45 8
14. System out println CORNERNt X value t Y value for int k 0 k nufCorners k System out println m k matrix CORNER Nt 4m k matrix X Nt 4m k matrix Y for int i 1 i lt nufCorners l i Om det r olika sorts h rn eller om i 1 ligger utanf r i s radie 2 s r de ej dubletter if m i 1 matrix CORNER m i matrix CORNER m i 1 matrix X m i matrix X LIMIT m isl1 matrix Xl m i matrix X LIMIT m i 1 matrix Y m i matrix Y LIMIT m i 1 matrix Y lt m i matrix Y LIMIT if m i matrix CORNER KONKAV retCorn 0 else if m i matrix CORNER KONVEX retCornt X else System out println Konstiga h rn else System out println Tog bort en dublett H rn Nr i Kollar f rsta och sista h rnen H rnen r ej dubletter LIMIT 2 if m nufCorners i matrix CORNER m 0 matrix CORNER m nufCorners 1 matrix X gt m 0 matrix X LIMIT 1 matrix X lt m 0 matrix X LIMIT m nufCorners 1 matrix Yl gt m 0 matrix Y LIMIT 1 matrix Y m 0 matrix Y LIMIT I System out println F rsta och sista h rn r EJ dubletter if m nufCorners 1 matrix CORNER KONKAV retCornt 0 else if m nufCorners 1 matrix CORNER KONVEX retCorn4 X T else System out println Konstiga h rn if m 0 matrix CORNER KONKAV retCornt 0 else if m 0 matrix
15. a new wall follow behaviour It will recursively follow all the remaining islands and make a complete exploration of the labyrinth Potential All labyrinths including any number of islands will be explored by the robot The algorithm strategy we needed to use would be Walk and follow wall Register continuously and draw the visual map Register areas with a wall or object we don t follow islands in a queue Identify current walls starting point and stop Remove the positions that has been explored already from the queue Go to the remaining areas in the queue and restart the algorithm QV Lpn PS LA FA rs This heuristic plan and algorithm would recursively explore any labyrinth 3 4 Implementation phase Collision avoidance behaviour First we had to start with the module that would give the robot a collision avoidance behaviour Here we simply checked at which proximity sensor values the robot ought to turn and at what speed so it could move along without hitting any objects or walls that may exist in the environment This was done quite easily and we got a grasp of the sensitivity of the proximity sensors and the values the walls would reflect from the infrared sensors The robot now was able to wander around randomly in the environment Wall following behaviour Later we wanted to implement the follow wall behaviour We let the left 90 degree sensor adjust itself towards a specific value that would make the ro
16. ate static final int NufUnitsInCirc int OMKRETS MOVEUNIT hjulenheter som g r i omkretsen Used when checking robot mode Khepera s omkrets i mm antal hela private private private private private static static static static static sensors final int SENS L90 0 final int SENS L45 1 final int SENS L10 2 final int SENS R10 3 final int SENS R45 4 final int SENS R90 5 final int SENS BR 6 final int SENS BL 7 static static static static static static static static final int T MOVING 0 final int T ON TARGET 1 final int M SPEED MODE 0 final int M POS MODE 1 final int M PWM MODE 0 int speedLeft 0 speedRight 0 int nufZero 0 16 int cruiseSpeed 10 boolean makeMap int positionLeft 0 positionRight 0 static OutputStream outputStream static InputStream inputStream Map map String originalCorners int mapWidth mapHeight int saved robotPos 20 h rn och konvex konkav XCoord YCoord angle per h rn 0 f r konkav h gersv ng och 1 f r konvex vanstersvang int cornerMatrix new int 20 4 int nufCorners static final int matrix CORNER 0 static final int matrix X 1 static final int matrix Y 2 static final int matrix ANGLE 3 static final int KONKAV 0 static final int KONVEX 1 Sensor limits for the corners static final int KONVEX LIMIT 1 static final int KONKAV LIMIT 1000 Konstuktorn public Robot int widt
17. bot move closely to the wall Here we encountered our first problem P1 The robot would have an uneven moving pattern and would jerk while it moved around This was a double problem since each jerk made the wheel counters loose accuracy and it was not nice to watch The cause was that we set the speed to 0 and changed the angle of the robot in each small turn We solved this by using a similar method as the Breitenberg cycle to be able to smoothly walk along the walls By adjusting the speed and turns gradually according to the values it had from the proximity sensors fig 2 we could finally get a smooth and accurate behaviour that would follow the walls at about 5 mm to 10 mm of distance In the wall following we had a few special cases to consider One was when the wall suddenly disappears and no more values can be read by the left 90 degree proximity sensor a convex turn We solved this by increasing the right wheel speed until the left 45 degree sensors could spot the wall again Wa IR proximity sensors Siemens SFH900 7 6 4 mm di 7 Fig 1 The 8 proximity sensors Now we encountered another problem P2 The robot wouldn t move into tight areas and would miss to turn We adjusted the front sensors so that they would not be as sensitive and by that allowing the robot to turn into tight areas as well It also encountered problems to move into v shaped walls We tried to adjust the sensors the same wa
18. e the reactive robot paradigm This decision was made when we realised that the sensors were useful and efficient enough to use this strategy We also had in mind from the lectures that this way may be easier then an integrated programming paradigm Parallel testing and programming of the robot was our main strategy to complete our task 2 Project description There were three levels of tasks that were included in this project The base goal was the task all groups were required to implement The main goal was the task definition of the project our group chose The additional task was the extra work we wanted to do with our robot in case we had enough time Base goal also called the introductory phase The introductory phase consists of forcing the robot to explore the environment and create its map You are free to choose the type of the map you wish your robot to use What is important is that your robot may handle simple labyrinths set up by an external observer You have to explain here the assumptions you make and make sure that you understand when they fail The next step is to let the robot locate itself using the map and to find the home position defined by a landmark possibly a bar code attached to the place Main goal To implement Hide and seek behaviour where the robot is supposed to look in a maze for a robot holding a light source on it Additional goal We had an idea to challenge another group in trying
19. eachedCorner false boolean leftTurn false rightTurn false Map localMap new Map mapWidth mapHeight int x 1 y 1 int newX newY reset L gger till 1 f r att kompensera f rsta h rnet som ej raknas if corners gt 0 corners t F lj vagg while corners 0 sens getProxSensors if leftTurn if sens SENS L45 KONVEX LIMIT ieftTurn true rightTurn false newX localMap getX newY localMap getY System out printin X newX4 NtY newY Ej dubletter om if mewX gt x newX lt x newY gt y mewy lt y I corners x newX y newY else System out println DUBLETT else if sens SENS L10 KONKAV LIMIT if rightTurn rightTurn true newX localMap getX newY localMap getY System out println X 4newX NtY 4newY Ej dubletter om if newX gt x newX lt x I newY gt y newY lt y corners x newX y 7 newY else System out println DUBLETT else rightTurn false else if sens SENS L451 600 amp amp sens SENS L90 gt 850 leftTurn false localMap updatePos getPos foliowWall sens emergencyBrake Staller sig vid f rsta b sta eg andra konkava h rn public void goToCorner int sens 20 int corners 0 boolean leftTurn false rightTurn false untilFoundWall adjust2Wall F61j vaag while corners lt 2 sens getProx
20. ent maps with higher resolutions Angle calculation Now the next step was to make sure that we knew the angle of the robot compared to its start angle Here we had to refresh our old trigonometric knowledge in order to come up with a correct formula We aimed to come up with a totally correct formula for any movements made by the robot and got a solution that was a bit too CPU demanding P3 We decided to approximate some distances to be straight and through this avoid a huge number of calculations that would give almost the same result as the approximation Constraint 2 Since it also would depend on the robots ability to not spin the wheels too much at some point relative to its countermovement and we were unsure of this factor we decided to go for the approximation The formula incremented the angle continuously and could give us the robots momentum angle at any time We calculated the distance the robot travelled at each movement by taking the average incremented value of left and right wheel distance and approximate that the robot went straight forward The momentum angle a of the robot was calculated by subtracting the total distance from left wheel dy with the total distance of the right wheel dr and see the proportion this total distance constitutes of a robot with the diameter of 55mm that performs a full 360 degree turn Thus the formula for a a 360 d dr 55 2x variables d total distance on left wheel d
21. ght sens SENS R10 DEGREE 10 1023f float diagRight sens SENS R45 DEGREE 45 1023f float sideRight sens SENS R90 DEGREE 90 1023f1 int speedR int cruiseSpeed sideLeft diagLeft frontLeft int speedL int cruiseSpeed sideRight diagRight frontRight skicka iv g hastigheterna setSpeed speedL speedR j Ett reaktivt v nster v gg f ljar beteende F ruts tter att roboten befinner sig n ra en v gg i b rjan Stannar n r den n tt x y i obs kartans koordinater origoX Map Width 2 och origoy Map Height 2 S public void followWall2Pos int x int y int sens booiean foundPos false boolean leftTurn false rightTurn false nufCorners 0 boolean isCorner false F lj vagg while foundPos sens getProxSensors map updatePos getPos if leftTurn Konvex h rn if sens SENS L45 KONVEX LIMIT leftTurn true rightTurn false cornerMatrix nufCorners matrix CORNER KONVEX cornerMatrix nufCorners matrix X map getX cornerMatrix nufCorners matrix Y map getY cornerMatrix nufCorners matrix ANGLE map getAngle nufCorners isCorner true Konkav h rn else if sens SENS L10 KONKAV LIMIT if rightTurn I rightTurn true cornerMatrix nufCorners matrix CORNER KONKAV cornerMatrix nufCorners matrix_X map getx cornerMatrix nufCorners matrix_Y map getyY cornerMatrix nufCorners matrix ANGLE map getAngle nufCorners
22. h int height mapWidth width mapHeight heignt map new Map mapWidth mapHeight originalCorners RRR R IRR ARERR RR RE PR RR KK eR HH HK RK IH IK TTR KR RR EE IR TK TRE ION TR RR RK RRR fe BEHAVIORS BERRA EAE e fie MC CN Ae ES EERRERREREREE e eee dee BR RR IR RRR EEK RK EERE RE REE KEK ES RAF d pp G fram t tills n got hinder hittas public void untilFoundWall int CLOSENESS 1000 int sens setSpeed cruiseSpeed cruiseSpeed boolean foundWall false while foundWall sens getProxSensors for int i 0 i sens length i 4 if sens i gt CLOSENESS foundWall true break Vrid roboten s att dess v nstra sida r parallell mot v ggen F ruts tter att roboten befinner sig n ra en v gg public void adjust2Wall int sens int i setSpeed cruiseSpeed cruiseSpeed boolean adjusted false while fadjusted adjusted true sens getProxSensors for i 1 i sens length i if sens SENS L90 lt sens i 300 adjusted false setSpeed 0 0 J V r reaktiva collision avoidiance behavior influerad av Braitenberg public void Braitenberg 17 float DEGREE 90 2f float DEGREE 45 4f float DEGREE 10 6f int sens getProxSensors float sideLeft sens SENS L90 DEGREE 90 1023f float diagLeft Sens SENS L45 DEGREE 45 1023f float frontLeft sens SENS L10 DEGREE 10 1023f float frontRi
23. he fastest route back to home position e To implement a walk from A to B behaviour e To add the measurement of the lengths of each wall so that we can have more information about the environment in order to record more unique data 12 8 Opinions Overall we found that the project was very interesting and stimulating But in order to achieve better results we wish to have had more time so that we could have planned each phase better The robot was adequate for its purpose to give us a good practical learning environment Maybe we wished for better accuracy of the sensors in some moments but this may be partly due to our own way of implementing the robot This course made us all very inspired and we would probably be interested in working with robots again The lectures and seminars have helped us a great deal to manage our results 9 References Lectures and seminars AI Robot course 125 HT 00 Jacek Malec 2000 Manuals Kephera Manual K Team 1998 Literature Mobile Robotics a practical introduction Chapter 5 Navigation Artificial Intelligence A modern approach S Russel P Norvig 1995 Seminar papers Intelligence without reason Rodney A Brooks 1991 Examination reports Object oriented robot control and real time simulation Tim Portnoff 1999 Object oriented robot simulator Henrik Lernmark 1999 Software RoboWorld Henrik Lermark Tim Portnoff 1999 Appendix A Java program class AI publ
24. ic static void main String args Window w new Window AI for Robots class Window extends Frame implements ActionListener Table table Button quitButton Button collisionAvoidButton Button followLightOrWall Button mapNPos Button localize Panel panel Robot kid Static int mapWidth 100 mapHeight 100 pubiic Window String s super s Frekk RK HR EIR HIM RAE KHIR RR RR RAKA F nster hantering stuff Za fe ek e eee e ede ke hee dede dee e dede de o ERK KK table new Table quitButton new Button Quit collisionAvoidButton new Button Collision Avoidance followLightOrWall new Button Follow Light mapNPos new Button Create Map localize new Button Find Home panel new Panel this setBounds 100 100 500 500 table setSize new Dimension 400 400 this add Center table panel add Center quitButton panel add Center collisionAvoidButton panel add Center followLightOrWall panel add Center mapNPos panel add Center localize this add South panei this pack this show quitButton addActionListener this collisionAvoidButton addActionListener this followLightOrWall addActionListener this mapNPos addActionListener this localize addActionListener this PRI RR RK KAR AAR KKK A Rw oe SLUT f nster hantering stuff APR K RRR KR ERK RR KK RR RR RK AAA EK int speedLeft speedRight kid new Robot mapW
25. idth mapHeight kid setupSerialConnection RRR ERERKRERK ERK EK RARE RKEREERHEKERERERK Z SJ LVA KODEN id BERRI KER ER HIM IK EERIE IKK HK ARA RAA EEE Collision Avoidiance 14 77 public void CA boolean quit false while true kid Braitenberg kid testZero Se Robot followWall2Pos public void mapAndPos int x y infile keyb new infile l if keyb Open System out printin Kan ej ppna tangentbordet System exit 0 kid new Robot mapWidth mapHeight kid setupSerialConnection kid setMapMaking true kid goToCorner kid reset System out printin SHTSHHBBHT A 4 SSTART AHERE RHR ERE HR E E kid followWall2Pos mapWidth 2 mapHeight 2 table setMap kid getModifiedMap table repaint kid setMapMaking false F ljer ljus men om det inte registrerar n got ljus s f ljer den v ggen public void followLightOrWall int sens boolean seeLight false boolean sawLight true while true sens kid getLightSensors for int i 0 i lt sens length 2 i if sens i lt 450 seeLight true if seeLight kid go2theLight sens else eg sawLight if sawLight kid untilFoundWall kid adjust2Wall else kid followWall kid getProxSensors sawLight seeLight seeLight false Se Robot localize public void localize kid localize
26. ity and values of the 8 infrared sensors proximity sensors Visual turrets visual graphical representation and sensitivity e Braitenberg control structure a method of robot control in an efficient smooth way 4 ROBOWORLD A simulation tool for the Kephera robot and its environment 3 2 Planning phase After achieving an understanding of the basics of the robot we synced our impressions of what was possible within the given frame of time We realised that Roboworld wouldn t be as useful as we first had thought The accuracy of Roboworld was good enough for basic understanding but not for real time programming feedback We came to these conclusions To use a reactive robot paradigm since we found that this would be easier to implement than a fully integrated system To focus on the base goal and the main goal meaning that we would focus on the navigation part and the graphical representation of the map Then we moved on to the follow light behaviour once the base goal was cleared We decided to not make any special coding in regard to the extended goal since this would be bad house holding with the limited time we had Once the first parts were ready we would bother about the extended goal later Our agreed agenda thus became A To focus on the base task B Complete the goal task C See how much time was left for extras 1 Use non lab time for read up on theory and come up with ideas to solve the implementation problems 2
27. lic boolean senseFront int frontLeft int frontRight int sensors getProxSensors if sensors 2 frontLeft sensors 3 frontRight return true else return false BER ERR RIKER ER EEK RIK R EK MHRA HR IIR RHI BK KKK AIR IH KH TRE ee KEIRA EKER ER REA e IME RR KEKE KARA 8 END METHODS FOR CONTROLLING KHEPERA VAR PESE E HOK R TR RR RR KR RR RR RR RA RR RR RA AR KER KKK RR RAR RK IL VAR doi LLL LE LLL E LEE LEE hh R8 eR eee d ehe dece OR e e ek ke ke dede ode dece k i doe END ROBOT IOK IOR IR RAA KRK AR RR RR RR RR RR KRA KAR RR RR EERE d BREE IR IR KHIR KRKA RII EKER IK IIHR HERRERA KK class Map gar int width height 1 ok nt 0 fritt 1 hinder anv nds ej int map boolean makeMap private static final int DIA 110 2 x Khepera s diameter i mm private static final float MOVEUNIT 0 08f Enhet som Khepera s hjul anvander i mm private static final float OMKRETS float Math PI DIA Omkrets 2 x kheperas i mm private static final float NufUnitsInCirc OMKRETS MOVEUNIT antal hjulenheter som i omkretsen private static final float k 55 2 MOVEUNIT Halva khepera i MOVEUNITS private static final float kDiag float k Math sqrt 2 Robotens koordinater i MOVEUNIT enheter float xCoord yCoord 26 double angle int oldLPos oldRPos int origoX origoY Undansparad info float saved xCoord saved yCoord double saved angle i
28. mf rt med initialvinkeln public int getAngle return int Math toDegrees angle Alt int Math rint xeturnerar robotens x koordinat i kartan public int getX return Math round xCoord 250 origoX returnerar robotens y koordinat i kartan public int getY return Math round yCoord 250 origoy returnerar robotens x koordinat i robotenheter public float getRealX return xCoord returnerar robotens y koordinat i robotenheter public float getRealY return yCoord returnerar true om robotenbefinner sig pa x y i kartkoordinater annars false F rsta fallet anv nds endast under map making kollar om man n tt m let eller om roboten har g tt ett helt varv X och Y koord r ter 0 0 och snurrat n stan helt varv B r nog ha st rre felmarginaler i alla fall i andra fallet Bo public boolean reachedPos int x int vil int LIMIT 3 System out printin X getxX Y getY if makeMap return getX origoX LIMIT amp amp getX lt origoX LIMIT amp amp getY gt origoY LIMIT amp amp getY lt origoY LIMIT amp amp getAngle gt 270 getAngie 270 else return getX x LIMIT amp amp getX lt x tLIMIT EE getY y LIMIT amp amp getY Xy LIMIT H public void print int x int xCoord 250 int y int yCoord 250 System out printin Robotens koora x yt System out printin
29. nt saved oldLPos saved oldRPos int saved origoX saved origoY public Map int x int y I xCoord 0 yCoord 0 oldLPos 0 oldRPos 0 angle 0 initierar kartan origoX x 2 origoY y 2 width x height y map new int width height for int i 0 i lt width i for int j 0 j lt height j map i j 1 public void saveContext saved xCoord xCoord saved yCoord yCoord saved angle angie saved oldLPos oldLPos saved oldRPos oldRPos saved origoX origoX saved origoY origoY public void restoreContext I xCoord saved xCoord yCoord saved yCoord angle saved angle oldLPos saved oldLPos oldRPos saved oldRPos origoX saved origoX origoY saved origoY public void updatePos int pos float distance distance traveled in MOVEUNIT int localL localR locaiL pos 0 oldLPos LocalR pos 1 oldRPos oldLPos pos 0 oldRPos pos 1 distance Math min localL localR distance localL localR 2f f rs ka ta minsta v rdet ist llet xCoord xCoord float Math sin angle distance yCoord yCoord float Math cos angle distance r kna ut ny vinkel anv nd gamla f rst angle 2 Math PI oldLPos oldRPos NufUnitsInCirc Satter robotens 1 8 punkter p kartan som ledig amp r det inte tillr cklig med 1 3 punkter public void set 1 float x y float beta float Math PI 4 angle 45grd alpha Mitten
30. ons and constraints P1 Jerky moving pattern causing inaccuracy Cause Speed of wheels and turns of the robot no synchronized Solution Adjusting speed and turns gradually and simultaneously inspired by Braitenberg cycle Well known algorithm for control of robot P2 Turning problems in tight labyrinths and v shaped walls Cause The front sensors were not set correctly according to what was needed Solution To use all of the front sensors and adjust them accordingly Constraint1 Avoid environments with v shaped walls P3 Exact calculation of angle and distance was too CPU demanding and not needed since the robots death reckoning deviation after further testing was expected too high Cause The robot will make numerous of small turns and movements and each movement will demand calculations Each turn will give some small deviation in the death reckoning Constraint2 Approximate turns to be straight and use the average distance value of the two wheels P4 Graphical representation mirrored compared to real world Cause Implementational mismatch Solution Correct the values so that they were displayed to show the real world P5 Identifying concave and convex corners Cause The sensors were not correctly tuned to adequately identify the corner types Solution Tuning them more adequately and use all relevant sensors in the process Constraint3 Not have too narrow convex and concave corners in the environment P6 Identifying co
31. orners matrix X localMap getX cornerMatrix nufCorners matrix Y localMap getY cornerMatrix nufCorners matrix ANGLE localMap getA ngle nufCorners isCorner true else if sens SENS L10 KONKAV LIMIT if rightTurn rightTurn true cornerMatrix nufCorners matrix CORNER KONKAV cornerMatrix nufCorners matrix X localMap getX cornerMatrix nufCorners matrix Y localMap getY cornerMatrix nufCorners matrix ANGLE 1ocalMap getAngle nufCorners isCorner true else rightTurn false else if sens SENS_L45 gt 600 amp amp sens SENS_L90 gt 850 leftTurn false followWall sens kollar om man n tt m let vid h rn if isCorner isCorner false if localMap reachedPos x y emergencyBrake foundPos true for int i 0 i lt nufCorners i System out print cornerMatrix i matrix CORNER System out printin newCorners corner Matrix2String cornerMatrix 19 System out println Original h rn originalCorners System out println Dessa h rn newCorners int cornerNr cornerNr compareCorners originalCorners newCorners if cornerNr 1 System out println G cornerNr h rn framat go2CornerNr cornerNr emergencyBrake restoreContext F rflyttar sig corners h rn fram F ruts tter att man befinner sig vid precis fore ett konkavt h rn public void go2CornerNr int corners int sens booiean r
32. ould be created for the port System out printin e getMessage System exit 0 catch UnsupportedCommOperationException e System out println The serial port could not be configured System out println e getMessage System exit 0 pubiic void frontalCollision sensibility int FLSens 900 int FRSens 900 25 int sensors getProxSensors if sensors SENS L10 FLSens sensors SENS R10 gt FRSens turnLeft 90 s l nge public void turnLeft int angle spara undan hastigheten int oldSpeedL speedLeft int oldSpeedR speedRight emergencyRrake vill eg bara s tta hast 0 Ber knar hur mycket varje hjul ska snurra int turnUnits int NufUnitsInCirc angle 360 Snurra hjulen setPosition 0 0 setReachPosition turnUnits turnUnits Kontollera om n tt positionen while reachedPos busy wait S tter roboten till ursprungshast setSpeed oldSpeedL oldSpeedR public void turnRight int angle spara undan hastigheten int oldSpeedL speedLeft int oldSpeedR speedRight emergencyBrake vill eg bara s tta hast 0 Beraknar hur mycket varje hjul ska snurra int turnUnits int NufUnitsInCirc angle 360 Snuxra hjulen setPosition 0 0 setReachPosition turnUnits turnUnits Kontollera om natt positionen while reachedPos busy wait S tter roboten till ursprungshast setSpeed oldSpeedL oldSpeedR pub
33. rners in symmetric labyrinths Cause The list of convex and concave corners may match another labyrinths list Solution Integrate current angle in the corner list future work Constraint4 Avoid symmetric labyrinths and symmetric corner list set ups P7 Conflict between light source and light on a wall Cause The sensor values of a wall with a light on would match the values of the light Constraint5 Only use a light source that come from straight above the environment 5 Results and the final status of the robot The status of the robot right now Base goal o Collision avoidance behaviour Works generally well and is included in the wall following behaviour But in some very narrow environments the robot will manoeuvre with a tight marginal only o Wall following behaviour Works well unless the walls are v shaped and labyrinth is too tight o Map drawing behaviour Draws a correct map in JAVA graphic representing the environment o Localisation behaviour Robot may be lifted and then finds home position without problems unless the labyrinth and corner set up is symmetric Main goal o Light source behaviour Finds a light source and follows wall alternatively without problems Additional goal Robot game Finds an opponent robot with a light source projected on top of it However not tested in reality since we lack an autonomous robot opponent Homing behaviour may be initialised by a click in interface or au
34. roximity values in each sector This was one of the ideas we had along with some kind of visual turret scans of the environment with built in landmarks But with little time left we decided to go for a more realistic and less time consuming approach where we recorded the complete series of convex and concave corners that exist in a random maze in a string We then lifted the robot and put it back at a random spot in the labyrinth The robot then followed walls again but this time only recorded the series of concave and convex corners This new string it matched against the old string which represented the original map set up When the comparation is done bit by bit until the robot knows how many corners it needs to move in order to arrive at the original home position We encountered a problem P5 with identifying these concave and convex corners consequently To fix this we had to tune the identification of corners better and not build too narrow labyrinths Constraint 3 The second problem P6 was that if the labyrinth consisted of a symmetric set up of concave and convex corners the matching would fail By avoiding these labyrinth we escaped the problem Constraint 4 We wouldn t have needed to make the robot move around the full labyrinth since only a few corners would have been enough to settle where the robot was relatively to the original map set up For total accuracy in finding home we decided to let it complete a full lap Light following
35. to track the opponents robot as fast as possible and then head for a predefined home goal This would be stimulating and force us to implement the robot more efficiently compare the AI checkers programs that needed to be optimally programmed to work well Ref AI game tournament in Al ground course 3 Continuous working process of the robot project 3 1 Introduction phase In our first experiments with the robot at the first lab occasion we closely studied what was possible to do with the robot and its possible potential through the available sensors We followed the Kephera user manual closely in order to not break anything and to learn the proper way to handle the robot The focus of our interest were 1 The physics of the robot identified the sensors LED indicators switches and the different parts of the robot Vision turtle wheels COM unit 2 The serial communication protocol to see the functionalities we were able to access through the serial cable connected to the robot 3 LabVIEW We familiarized us with the LabVIEW environment that enabled us to use the communication protocol through a graphic interface that could visualize the sensor values of the robot We could then through experimentation and with the help of information in the manual grasp the following concepts e Steering system of the robot Speed capabilities of the robot The accuracy of the wheels The accuracy and function of the wheel counters Capac
36. tomatically future work as soon as it finds the opponent 6 Final Analysis We find that the reactive model was a good choice since we could solve problems in an ad hoc way By dividing the complete task in different behaviours the work was simplified and less complex The divide and conquer strategy for the complete project was therefore a successful one If we would have done this project over again we would probably use the reactive model once again Knowing then that it is quite difficult to match the robot world with the real world due to uneven light conditions uneven sensor readings death reckoning deviations real time programming bugs and our own inexperience in robot programming we would set aside more work hours for the navigation part of the project We also found that it is very important to plan each step ahead so that the time is used as effectively possible We would probably have avoided the fact that we didn t have any practical use of the graphical representation of our robot at all if we had discussed this in the planning phase 7 Future work on the robot We would want to implement the following in an improved version of our robot e Acomplete heuristic strategy so that the robot may explore all possible labyrinths including islands e A test of the generated map to see if the map contains a symmetric sequence of convex and concave corners e To implement a follow wall routine with right walls so that it may find t
37. y getReply reply reply substring reply indexOf prefix prefix length 1 Anv nd en tokenizer f r att separera substr ngarna med v rdena StringTokenizer token new StringTokenizer reply n r f int result new int token countTokens Parsa substr ngarna med v rdena for int i 0 i lt result length i String siffror token nextToken try 4 result i Integer parseInt siffror catch Exception e System out println Warning Problem parsing values System out println String to parse reply System out printlin e return result Robert Pallbo s Initierar seriekommunikationen public static void setupSerialConnection CommPortIdentifier portId SerialPort serialPort ppna dev modem try portId CommPortIdentifier getPortIdentifier dev modem serialPort SerialPort portId open KheperaController 2000 outputStream serialPort getOutputStream inputStream serialPort getiInputStream serialPort setSerialPortParams 19200 SerialPort DATABITS 8 SerialPort STOPBITS 2 SerialPort PARITY NONE catch NoSuchPortException e System out println No such port System out printin e getMessage System exit 0 catch PortinUseException e System out println The port is already open System out println e getMessage System exit 0 catch IOException e System out printin No output or input stream c
38. y as we did with the tight turn problem above but this would give problems in other parts of the robots total behaviour Therefore we decided to not let the environment consist of any v shaped angles Constraint 1 Visual map representation Now when the wall following behaviour worked we wanted to focus on the visual representation Some tests were conducted to see whether the death reckoning strategy based on the wheel counters and the angle of the robot would suffice to keep track of where the robot was at any point We believed that this together with the proximity sensors to correct for eventual deviations would be accurate enough We decided to first represent it as basic as possible and therefore chose ASCII symbols in a grid of 100x100 to represent the physical world Each grid square would be 2x2 cm We chose the following map legends to represent different areas in the map Unknown ground which means the whole map grid consisted of in the beginning 0 Free ground where the robot know there are free positions in the map Where the current wall is positioned on the map Then our plan was to let each new object or wall a potential island have 2 3 4 and so on Future work After the basic map was completed we decided to use JAVA graphics to represent the map and an interactive interface for controlling the robot more easily We found it more user friendly and also had options to pres
Download Pdf Manuals
Related Search
AI robotprojektreport