Real-time model predictive obstacle avoidance control for vehicles with reduced computational effort using constraints of prohibited region

Abstract

This paper focuses on a real-time obstacle avoidance control method for vehicles using model predictive control (MPC). MPC can optimize the motion of the vehicle over a finite time horizon while satisfying various constraints such as vehicle dynamics, the road width and the steering range. However, the computational cost is too large for conducting real-time control. In this paper, a collision avoidance is realized by MPC with constraints for avoiding prohibited regions represented as circles. We approximate this region into a half plane separated by the tangent of the prohibited region. By handling approximated regions as constraints of the road width of MPC, we can implement the collision avoidance algorithm into the controller without increasing the computational cost. Moreover, in order to reduce the computational effort, we transform the nonlinear vehicle dynamics into reduced order and linearizable subsystems called time-state control form (TSCF). The effectiveness of the proposed method is proved by comparative simulations with conventional method where artificial potential method is applied to MPC. In addition, we conduct two experiments using a 1/10 scale vehicle which is equipped with a laser range finder to execute obstacle detection and localization. We show that real-time control can be realized even if we use an on-board embedded CPU which runs at the frequency of 500MHz.