AUTONOMOUS MAP LEARNING FOR A MULTI- SENSOR MOBILE ROBOT USING DIKTIOMETRIC REPRESENTATION AND NEGOTIATION MECHANISM

In this paper we present a method for a multi-sensor mobile robot to explore autonomously an unknown environment. The method tries to solve the main problems involved by such a task. In particular we present a model to represent geographic knowledge, based on an extension of the "Diktiometric representation" of Engelson and McDermott (1). We paid special attention to the maintenance of this model, providing mechanism to allow the consistent fusion of sensory observation. Furthermore we argue that, due to the different capabilities of the devices of a multi-sensor system, the only interaction between the sensors should be indirect and based on the individual effect that each sensor has on the system controller. Therefore we present a negotiation mechanism allowing to integrate the world knowledge represented in different models, each of which is updated with the sensory information provided by a specific devices. In our work this integration involves only the exploration strategies of each representation.