Next: The Reinforcement Learning Algorithm Up: Reinforcement Learning Previous: Reinforcement Learning
The task confronting the Learning Agent is to choose actions (for both motion and vision) in order to reach the current target location while minimizing the use of the camera. The Map Manager informs the Learning Agent about the target location. If the robot becomes blocked, the Rescuer will ask the Map Manager for a new target (a diverting target), and then the Learning Agent will take control and choose actions to reach that new target. Once the diverting target is reached, the Rescuer may be able to set the current target to be the original goal, and then the Learning Agent will attempt to move to that target (and hence, solve the original task).