Neural networks for autonomous robot navigation

摘要

Research at West Virginia University directed at developing autonomous mobile robot navigation systems for use in underground mines is discussed. One of the focal points has been the exploration of applications for neural network technology in the navigation context. An overview of the various experiments as they relate to each other is given along with some of the conclusions drawn about neural network applications resulting from this work. The networks have excelled in applications involving pattern recognition such as the neural area recognition system, the landmark recognition system, and the noise distribution function estimator, in these applications the difficulties encountered in configuring the networks by trial and error and the slow network training process were outweighed by the performance the networks eventually provided in difficult signal processing tasks. In other applications such as the early steering control system and the neural position estimator the performance of the networks has proven inferior to what is believed could have been achieved more easily by conventional means.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>