Distance Sensor Data Integration and Prediction

In this paper we describe an approach to sensor data integration in order to obtain a qualitative and robust interpretation of the robot environment, which could be used as input in the qualitative robot navigation algorithms developed by our group. This approach consists of: obtaining patterns of distance zones from the sensor readings; comparing statically these patterns in order to detect non-working sensors; integrating the patterns obtained by each kind of sensor in order to obtain a final pattern which detects obstacles of any sort; predicting how the distance pattern will be influenced by robot motion and comparing this prediction with reality in order to detect incoherent sensorimotor situations. This approach has been applied to a real Pioneer 2 robot and promising results are obtained.