Car cruising system incorporating ART-FNN man-machine interface in adaptation to various driving behaviors

This research suggests a new approach to Man-Machine interface, that is bas ed on the characteristics of an imaginary individual's driving behavior and the easily detected car control figures. The adaptive resonance theory (AR T) network is combined with the computer-aided learning of the Fuzzy Neural Network to obtain the different appropriate safe car-following headway bet ween cars that are running under different speeds. The framework of this sy stem is divided into two major parts. The first part uses the ART network, which uses the differences between the required least horizontal braking di stance of a car, and the distance between the car and the obstacle standing on the end of the lane at the moment that the driver applied the brakes, d uring a test where the car is running on a straight lane at accelerated spe ed. An off-line Self-organizing cluster discovering is conducted. The clust ers of the above-mentioned two distances obtained through the ART network a re used to determine the levels of the risk-taking factors of the drivers. Then the output of this ART network (i.e. risk-taking factor level), car sp eed and free travel of brake pedal are used as the fuzzy neural network (FN N) input, the appropriate car-following headway serves as the FNN output. T hrough the learning capability of artificial neural network, the complex me mbership functions between the inputs and the output can be efficiently est ablished. Finally, the appropriate car-following headway predicted by the F NN are compared with the actual field data to prove the accuracy of the sys tem in predicting the different safe car-following headway, given different driver and different car characteristics driven at different speeds. The c ompletion of this system provides a feasible course for the development of a neural network in an individually-oriented man-machine interface system.