Towards longitudinal and lateral coupling control of autonomous vehicles using offset free MPC

Model predictive control (MPC) is widely used in the motion control of autonomous vehicles. However, the conventional MPC relies on an accurate model and cannot achieve offset-free tracking in steady-state when the autonomous vehicle encounters external disturbance and parameter perturbation. Based on our recently designed C++ offset-free MPC (OF-MPC) solver, this paper uses MPC to solve the longitudinal and lateral coupling control problem directly. In the modeling stage, the constraints of the tire friction circle are approximated by the polygon. And then the polygon constraints are mapped into the state and control space. Finally, the tire friction circle constraints are approximated as convex linear inequality constraints, which are used as the vehicle stability boundary condition. Meanwhile, a convex optimization method is used to solve the reference value problem of MPC, and a systematic method that can prevent the reference value from being outside the feasible region is proposed. The real-time simulation and actual vehicle results show that the algorithm can solve both lateral and longitudinal disturbances problems. The algorithm achieves better tracking accuracy and high-speed stability due to consideration of the longitudinal and lateral coupling constraints. Benefiting from the proposed efficient solver, the average time to complete all calculations, including logic processing and input–output processing, is only about 5ms when the horizon length is 50.