Integration Stability Analysis Of Nonlinear Model Predictive Motion Control For Autonomous Vehicles

Abstract

There has been a growing interest in nonlinear model predictive control (NMPC) for the motion control of autonomous driving. However, it is odd that integration stability has not been considered enough when developing these applications for autonomous driving. The stabilized explicit RungeKutta (RK) integration method is proposed in this paper to solve the integration stability problem in motion control of autonomous vehicles (AVs). In comparison with other explicit integration methods, this method provides a wider stable region and is more efficient than implicit integration methods. This integration method is integrated into the framework of offset free nonlinear MPC solver based on GRAMPC by us. As a result, the problem of computational stability at low speeds of motion control can be resolved. At the same time, a larger integration step size can be adopted, and nonlinear MPC becomes more computationally efficient. The results of simulation and real vehicle experiment results show that the problem of low-speed integration stability when using non-linear dynamic model as a prediction model is successfully solved, which has been ignored in many previous studies.