An autonomous robot interacting with people in the real world is faced with a large amount of sensory data and uncertainty in its action outcomes. A robot requested from humans to serve has to make plans to achieve the goals; it needs to decide on seq...
An autonomous robot interacting with people in the real world is faced with a large amount of sensory data and uncertainty in its action outcomes. A robot requested from humans to serve has to make plans to achieve the goals; it needs to decide on sequences of actions and robustly run behaviors even unpredictable situations. While a robot has the information about environments and elaborately accomplish its goals in conventional systems, domains for a robot are aggrandized in real world and it requires autonomous behaviors and achievement of goals. Hybrid systems between reactive and deliberative make the needs possible but it is difficult to decide on the rate of two systems within hybrid system.In this paper, we propose the hybrid system with behavior network and planning for generating autonomous behaviors and achieving goals. Whereas Behavior networks are apt to solve reactive problems with several goals, they are rarely applied to problems with complex plans or inference. In addition the more complex situations a robot confronts, the more nodes of the behavior network the designer considers. We decompose global goals of a mobile robot and construct sequences of modules with sub-goals from the priority in each commands. It helps a robot to quickly react in dynamic situations as well as achieve goals using proposed architecture.We demonstrate our module planner with behavior network modules on the Webot simulator and KheperaII real robot. The task is the delivery of objects from one room to another room when users give a robot commands to deliver an object. A robot may confront dynamic situations such as that a door for entering the room is closed or moving obstacles exist. Experimental results with various situations have shown that a robot can achieve goals and make dynamic module sequences albeit there are unpredictable situations like closed doors or moving obstacles.