抄録
In this paper, We discuss a method of decision making of a behavior-based robot with some modules which decide robot's action. Subsumption Architecture(SA) makes the modules a layered structure, and upper layer module's action is employed in priority to lower layer module. Such a easy adoption way gives easy implementation but has the problem that lots of inefficient actions are employed. This problem is caused by employing upper module regardless of other modules. Then, We find solutions to this problem by the method to represent actions by Potential Function(PF). This method adopt the action which collects maximum votes from modules. With event driven state transition, robot makes decision of its action with appropriate set of modules which is changed depending on situation. We apply this method to navigation tasks in some corridor environments and show simulation results.
