Vehicle longitudinal control using throttles and brakes

Abstract

One of the main challenges of automotive vehicle longitudinal control is integrating throttle and brake control. The difficulties are mainly due to the necessity of switching throttles and brakes, and the non-symmetry of time delays - large delay for brakes, and small delay for throttles. Existing control techniques switch between throttles and brakes based on a simple predetermined criterion. As a result, switchings are frequent, and ride jerky. In this paper, we propose a control strategy where two control laws, one for the throttle and one for the brake, are computed simultaneously to optimize a certain tracking criterion by a learning algorithm. The two computed control signals are then used to determine whether to activate the throttle, or the brake, or inaction. The resulting control law is shown to give dramatically smoother behaviors for vehicle tracking.