A Neurosymbolic Hybrid Approach for Landmark Recognition and Robot Localization

Robot self localization is a crucial issue in autonomous ro- botic research. In the last years, several approaches have been proposed to solve this problem. In this paper, we describe a landmark based neu- rosymbolic hybrid approach to tackle the global localization problem.We use the same approach to cope with the whole problem: from landmark recognition to position estimation. The map given to the robot is inter- preted by a neurosymbolic system (formed by a weightless neural network and a BDI agent) for extracting landmark information. A \virtual neural sensor" is used, during robot navigation, for detecting the landmarks in the real environment. These information (map and detected landmarks) are ¯nally processed by a uni¯ed neurosymbolic hybrid system (NSP) for determining the robot location on the given map.