Application Of Neurosymbolic Integration For Environment Modelling In Mobile Robots

We present an architecture for representing spatial information on autonomous robots. This architecture integrates several kinds of representations each of which is tailored for different uses by the robot control software. We discuss various issues regarding neurosymbolic integration within this architecture. For one particular problem-extracting topological information from metric occupancy maps-various methods for their solution have been evaluated. Preliminary empirical results based on our current implementation are given.