EVOLUTION OF AN ARTIFICIAL NEURAL-NETWORK-BASED AUTONOMOUS LAND VEHICLE CONTROLLER

This paper presents an evolutionary method for creating an artificial neural network based autonomous land vehicle controller, The evolved c ontrollers perform better in unseen situations than those trained with an error backpropagation learning algorithm designed for this task, I n this paper, an overview of the previous connectionist based approach es to this task is given, and the evolutionary algorithms used in this study are described in detail, Methods for reducing the high computat ional costs of training artificial neural networks with evolutionary a lgorithms are explored, Error metrics specific to the task of autonomo us vehicle control are introduced; the evolutionary algorithms guided by these error metrics reveal improved performance over those guided b y the standard sum-squared error metric. Finally, techniques for integ rating evolutionary search and error backpropagation are presented, Th e evolved networks are designed to control Carnegie Mellon University' s NAVLAB vehicles in road following tasks.