Self-tuning of safety-assistance brake controller of wheeled vehicles

Vehicle active safety systems can assist the driver to retain control of the vehicle in situations where the vehicle might lose control. This paper considers achieving vehicle active safety by employing a safety-assistance brake controller to perform anti-lock control, brake-based traction control, yaw-stability control and brake-force distribution. Main difficulties in synthesizing the safety-assistance brake controller are that the vehicle-tire system is a complicated nonlinear dynamic process, and active safety demands several control objectives that may conflict with each other. We adjoin the control objectives in a single cost function; then employ a self-tuning algorithm to optimize the control policy. The safety-assistance brake controller is composed of a radial basis function network; the self-tuning algorithm uses adaptive critics to guide optimization. Adaptive critics refer to a nominal vehicle-tire model; however training examples cover entirely the admissible domain of operations of the braking control system. The obtained safety-assistance brake controller is investigated for several vehicles by simulations. Results compare critical tests of the vehicles with and without active safety brake controls. It is shown that the self-tuning algorithm is effective in obtaining a useful safety-assistance brake controller.