Control robot by multiagent control architecture

In this paper we propose a behavior-based architecture, which can also be seen as a multi-agent architecture, to control a situated and embodied autonomous mobile robot, under a close sensing action relationship. Preexisting knowledge can play a meta-level role, being used in order to provide an adviser capability which is combined with the basic reactive feature of the robot. One of the challenges of the behavior-based approach is the integration and coordination of multiple different behaviors. This problem was faced using combination operators applied to lower level behaviors and producing useful and coherent higher level behaviors. Our autonomous mobile robot also exhibits learning capabilities, namely the ability to learn new behaviors in an autonomous way. In this context it was designed to the elementary behaviors a connectionist architecture that encompasses the capabilities of building specific basic competences. Through the exploration of the proposed control architecture a mobile platform has become an useful autonomous mobile robot It is able to deal with dynamic environments, evolving in a coherent, relevant and adequate way according to both the current situation as well as with its assigned missions.