Neural-network synthesis of integrated brake controller of electrical vehicle

The integrated brake controller involves in one electronic unit the antilock control, traction control, yaw stability control and braking force distribution that manipulate forces and timing of braking on each wheel. Main difficulty is that the vehicle-tire dynamics is a nonlinear process. Secondly, controls of antilock, traction, yaw stability and braking force distribution operate for different objectives. Therefore, the integrated brake controller has to satisfy multiple objectives subject to constraints imposed by the nonlinear vehicle-tire dynamics. This paper presents a method of reinforcement synthesis to automatically synthesize an integrated brake control policy. The control objectives are expressed by a cost function and the reinforcement synthesis algorithm adapts the controller, a radial basis function network (RBFN), to minimize the cost. With sufficient training, the RBFN would be optimized for performing integrated brake control. The obtained controller is investigated for critical tests of a vehicle simulating on CarSim. Results show that reinforcement synthesis can work out a useful integrated brake controller.