| 摘要 |
PURPOSE:To obviate tremendous algorithm structure in a conventional system and simplify system structure by making the system learn vehicle motion by means of a neural net and totally controlling a vehicle depending upon the result. CONSTITUTION:To a neural net A, the results of vehicle motion, i.e., yaw angle speed, rolling angle speed, pitch angle speed, lengthwise acceleration, lateral acceleration vertical acceleration and vehicle speed are input. Then, the neural net A outputs vehicle control parameters, i.e., front wheel steering angle, rear wheel steering angle, throttle opening degree, fuel injection quantity, ignition time, cylinder pressure and brake pressure. Whether the output vehicle control parameters coincide with teacher signals, which are actual experimental values, or not is verified. As a result, if the calculated values do not coincide with teacher signals, the weight of the neural net A is corrected so that the differences between their values may be less than permissible levels. After that, actuators are made to operate by outputting corrected ideal control signals to control vehicle motion characteristics. |
