Abstract
In this paper, neural network (NN) theory and its genetic algorithm (GA) learning are applied to steering control of cars. It is difficult to acquire steering operation on various curved roads. The NN controller consists of 3 NNs. It uses look ahead information on straight roads and loose curved roads and uses car motion information too on tight curved roads at high speed. The controller uses human like information by using the 3 NNs structure. The controller steers successfully on various curved roads as a result of the GA learning. The efficiency and feasibility of this method is verified by numerical simulation.