A connectionist approach to the acquisition of elementary navigation competencies by a mobile robot

Abstract

The acquisition of navigation competencies by a mobile robot is one of the key problems in the development of an autonomous mobile robot. This subject has been approached in the scope of the development of ARCoS-autonomous mobile robot control system-whose purpose is to transform a mobile platform in a real autonomous mobile robot able to deal with several tasks in different environments and also able to evolve through the dynamic acquisition of new navigation competencies. In our approach the robot competencies will emerge from the reactive agent's actions involved in the assigned mission. Each reactive agent (RA) is responsible for control the mobile robot in the achievement of an elementary competence. The proposed RA architecture, integrating the connectionist and symbolic paradigms, aims to be rich enough to accommodate agents whose competence is easily described and programmed as well as those which are hard, boring or even impossible to be specified being learned from the experience. The RA architecture also implements the associated cooperation strategies with the other agents participating in the execution of the same task and therefore, allowing the mobile robot to exhibit more complex competence.