The Research On Application Of Nonlinear Model Predictive Control Algorithm In Unmanned Vehicle Trajectory Tracking

Autonomous driving has been a research hotspot in recent years,and a high-performance tracking algorithm that can adapt to changes in vehicle speed is one of the keys to ensure the safe driving of unmanned vehicles.Aiming at the vehicle following control function in automatic driving,this paper proposes an improved Nonlinear Model Predictive Control(NMPC)path tracking algorithm,and on this basis,proposes a lateral and longitudinal coupled trajectory tracking algorithm.The algorithm’s target lateral and longitudinal control accuracy in real vehicle experiments are 0.7 m and 1 m,respectively.A simulation comparison study of three path tracking algorithms was carried out,and based on the comparative analysis,an improved NMPC path tracking algorithm was proposed.The simulation condition is a uniform circular condition,and the three algorithms are Linear Quadratic Regulator(LQR),Linear Model Predictive Control(LMPC)and NMPC.The simulation results show that the lateral and longitudinal tracking performance of NMPC is better than that of LQR and LMPC in low and medium speed conditions.However,under medium and high speed conditions,NMPC cannot complete the path tracking task.In order to solve the above problem,the NMPC path tracking algorithm is improved,and the lateral steadystate error of the improved algorithm at medium and high speeds is 0.12 m.A trajectory tracking algorithm with lateral and longitudinal coupling was proposed,which solves the problem that the path tracking algorithm cannot cope with variable speed conditions well.On the basis of the proposed improved NMPC path tracking algorithm,the algorithm takes the speed as one of the control variables.A simulation and comparison study of the lateral and longitudinal decoupled and coupled trajectory tracking algorithms based on NMPC was carried out in the two scenarios of lane change and merging in the autonomous driving lane and off-ramp.The simulation results show that the lateral and longitudinal tracking performance of the trajectory tracking algorithm based on the lateral and longitudinal coupling is better than that of the trajectory tracking algorithm based on the lateral and longitudinal decoupling.Aiming at the experimental vehicle that only supports steering wheel torque input,a "feedforward + feedback" steering control strategy was proposed,which can accurately convert the rotation angle command calculated by the tracking algorithm into a torque command.The steering system dynamics model of the unmanned vehicle is established,and the key parameters are identified by the least recursive square method.The feedforward torque value is calculated by the established dynamic equation,the feedback torque value is calculated by the difference between the actual rotation angle and the desired rotation angle,and the sum of the feedforward and feedback is taken as the final control command.In the simulation experiment,the maximum turning angle following error is 5.3 degrees;in the actual vehicle experiment,it is 6.1 degrees.Both the simulation and the actual vehicle experiment results prove the feasibility of the steering control strategy.The NMPC-based lateral and longitudinal coupled trajectory tracking algorithm is tested in real vehicles in the scenes of lane change and merging and off-ramp in the autonomous driving lane.The lateral error peak in the scene of lane change in the autonomous driving lane is 0.17 m,the longitudinal error peak is-0.41 m,the lateral error peak in the ramp-off scene is-0.54 m,and the longitudinal error peak is-0.82 m.The results show that the control accuracy of the trajectory tracking algorithm meets the requirements.It provides a theoretical and experimental basis for the application of the NMPC algorithm in the high-degree-of-freedom vehicle dynamics model in the future.