1. CS 1 with Robots Behavior Based Control Institute for Personal Robots in Education (IPRE)‏

2. Aug 29 2007 2 Behavior Based Control Behavior 1 Behavior 2 Behavior 3 Arbitrator Robot Actions

3. Aug 29 2007 3 Behavior Based Control Behavior based control allows you (the programmer) to specify robot behaviors. Multiple behaviors “fight it out” for control of the robot.

4. Aug 29 2007 4 Each Behavior May use sensor values to determine what to ask the robot to do, or may not. Makes a recommendation to the arbitrator about what action the robot should take. or....tells the arbitrator that it doesn't have an opinion.

5. Aug 29 2007 5 The Arbitrator Asks each behavior what the robot should do. Decides which behavior(s) to use. Advanced Arbitrators may even allow the robot to do a mixture of different behaviors at the same time!

6. Aug 29 2007 6 Implementing a Behavior based control system You need to specify each behavior...and... An arbitrator that decides which behavior to follow at any particular point in time and makes the robot's actuators follow that behavior's recommendation.

7. Aug 29 2007 7 Problem Statement: Your robot should move continuously looking for a bright light. When it finds a bright light, it should go towards it. The robot should turn to avoid hitting obstacles.

8. Aug 29 2007 8 Finding Behaviors (look for verbs!)‏ Your robot should move continuously looking for a bright light. When it finds a bright light, it should go towards it. The robot should turn to avoid hitting obstacles.

9. Aug 29 2007 9 Summarized: Move Look (for light)‏ Go to the light Avoid obstacles

10. Aug 29 2007 10 Combined/Simplified: Move Go towards light Avoid obstacles

11. Aug 29 2007 11 Prioritized: 1st: Avoid obstacles 2nd: Go towards light 3rd: Move

12. Aug 29 2007 12 Questions about behaviors How does the behavior communicate with the arbitrator? How does it indicate what action it wants the robot to take? How does it indicate that it doesn't have a recommendation? All of these questions can be answered by building the Arbitrator.

13. Aug 29 2007 13 The Arbitrator Calls each behavior. Sees what each behavior wants to do. Selects the recommendation from the highest priority behavior. Tells the robot to do that behavior's recommended action.

14. Aug 29 2007 14 The Arbitrator/Behavior interface Each behavior needs to return the same data (in the same order) to the Arbitrator. You can pick any data and any order you want, but you must be consistent in all of your code. For this example, we choose to return: Boolean – Tells if a behavior has a recommendation (True) or not (False)‏ Translate Value – Float from -1.0 to 1.0 telling the recommended translate amount. (0.0 is stopped)‏ Rotate Value – Float from -1.0 to 1.0 telling the recommended rotate amount. (0.0 is no rotation)‏ [ True, 0.0, 0.0 ] - Represents a recommendation to stop completely.

15. Aug 29 2007 15 Behavior 1: Cruse #Move behavior: Always recommends that the robot move. def move(): return( [ True, -1.0, 0.0 ] )‏

16. Aug 29 2007 16 Behavior 2: Avoid Obstacles #Avoid behavior, makes the robot turn away from obstacles. def avoid(): sensorVals = getIR()‏ leftV = sensorVals[0] rightV = sensorVals[1] if (leftV == 0) or (rightV == 0): return( [ True, 0.0, 1.0 ] )‏ else: return( [ False, 0.0, 0.0 ] )‏

17. Aug 29 2007 17 The Arbitrator: Putting Avoid & Move together! #Calls each behavior in the order listed. If a behavior gives a recommendation, # arbitrate passes the recommended action on to the robot and then skips the # rest of the behaviors. (So avoid has a higher priority than move.)‏ def arbitrate(): behaviors = [avoid, move] for behavior in behaviors: values = behavior()‏ hasRec = values[0] transVal = values[1] rotVal = values[2] if (hasRec == True): translate(transVal)‏ rotate(rotVal)‏ return()‏

18. Aug 29 2007 18 Behavior 3: Seek Light #seekLight behavior: turns towards (bright) light def seekLight(): values = getLight()‏ leftV = values[0] rightV = values[2] if ( leftV > 200) and (rightV > 200): return( [False, 0.0, 0.0] )‏ else: if (leftV < rightV): return( [True, -0.5, 0.33] )‏ else: return( [True, -0.5, -0.33] )‏

19. Aug 29 2007 19 Updating the Arbitrator for a new behavior #Calls each behavior in the order listed. If a behavior gives a recommendation, # arbitrate passes the recommended action on to the robot and then skips the # rest of the behaviors. (So avoid has a higher priority than move.)‏ def arbitrate(): behaviors = [avoid, seekLight, move] for behavior in behaviors: values = behavior()‏ hasRec = values[0] transVal = values[1] rotVal = values[2] if (hasRec == True): translate(transVal)‏ rotate(rotVal)‏ return()‏