Efficient learning of variable-resolution cognitive maps for autonomous indoor navigation

This paper presents an adaptive method that allows mobile robots to learn c ognitive maps of indoor environments incrementally and on-line. Our approac h models the environment by means of a variable-resolution partitioning tha t discretizes the world in perceptually homogeneous regions. The resulting model incorporates both a compact geometrical representation of the environ ment and a topological map of the spatial relationships between its obstacl e-free areas. The efficiency of the learning process is based on the use of local memory-based techniques for partitioning and of active learning tech niques for selecting the most appropriate region to be explored nest. In ad dition, a feed-forward neural network is used to interpret sensor readings. We present experimental results obtained with two different mobile robots, namely a Nomad 200 and a Khepera. The current implementation of the method relies on the assumption that obstacles are parallel or perpendicular to e ach other. This results in variable-resolution partitionings consisting of simple rectangular partitions and reduces the complexity of treating the un derlying geometrical properties.