Neural network based landmark recognition for robot navigation

Abstract

The problem of landmark recognition is essential to mobile robot navigation since the landmarks can give information about the global position of the robot as well as information about local traffic conditions. The authors describe a robust landmark technique based on reconfigurable neural networks. They present a brief list of current landmark recognition and mobile robot guidance techniques and justify the choice of the neural network model. Learning rules with update normalization that improve learning stability are introduced. The learning rates of this network when trained with actual landmarks are reported.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>