BUILDING A MAP FOR ROBOT NAVIGATION USING A THEORY OF COGNITIVE MAPS

In this paper we describe the application of our theory of cognitive maps [1] to the problem of building a map for a navigating robot equipped with sonar sensors. At the core of this theory is the notion that a representation is computed for each local space the robot visits. These representations, connected in the way they are experienced, form the robot’s cognitive map. Real world implementations which compute representations of a robot’s environment face the problem of errors due to inaccuracies in the sensory data and errors in the robot’s location due to wheel slippage. Sonar sensing devices have the added problem that there is not much sensory data to begin with. We show that using our theory of cognitive maps it is possible to build a reasonable representation of the robot’s environment without complex error correction procedures.