Automatic tuning method of membership functions in simulation of driving control of a model car

We carry out the tuning of the membership functions using the learning procedure for studying neural networks. The tuning method is applied to the simulation of driving control of a model car. The 3/spl times/3 (=9) control rules work well for the driving simulation of the car to run on a straight road. There are some rules which are hardly used when the number of the rules increases up to 7/spl times/7 (=49). After tuning of membership functions, the simulation results give similar trajectories of the car in both 3/spl times/3 and 7/spl times/7 type rules.