Overtaking control strategy based on model predictive control with varying horizon for unmanned ground vehicle

Abstract

In this paper, a synthesized novel strategy of varying predictive horizon-based model predictive control is proposed for the overtaking control of unmanned ground vehicle. The whole control strategy includes path planning and path tracking. First, the preferred path in presence of diverse constraints of states, inputs, and collision avoidance can be calculated using Gauss pseudospectral method where expected position, velocity, and attitude are provided. Correspondingly, the continuous optimal control problem is converted to discrete nonlinear programming. Second, model predictive control is developed for tracking the optimized path. Considering the effect of the predictive horizon and the Gauss points’ distribution on tracking performance, the varying predictive horizon is introduced to improve the tracking accuracy in non-smooth path. By the varying predictive horizon-based model predictive control method, less computation burden and better control performance are achieved. For the difference between the mathematical expressions and the real unmanned ground vehicle dynamics, genetic algorithm is utilized to identify the parameters of tire model. Simulations in MATLAB and CarSim are both implemented to illustrate the effectiveness of the proposed method.