Research On Autonomous Vehicle Motion Control Algorithm Based On Nonlinear Model Predictive Control

In recent years,rapid advancements in computer technology,chip technology,communication technology,and sensing technology have significantly enhanced the feasibility of autonomous driving.Autonomous driving technology primarily consists of three major modules:perception and localization,decision-making and planning,and motion control.Motion control,as one of the key technologies in autonomous driving,is responsible for precisely and stably tracking the reference trajectories provided by the upper-level decision-making and planning modules.Its control performance directly affects vehicle safety and user experience during driving.Among numerous motion control methods,Nonlinear Model Predictive Control(NMPC)has been widely applied to autonomous driving motion control due to its systematic consideration of model nonlinearity and its absolute advantages in handling optimal control problems with multiple constraints.Although NMPC-based autonomous driving motion control technology has made significant progress,there are still critical issues that need to be addressed,especially in the field of high-level autonomous driving.These issues include:(1)the existing vehicle motion control model has limited applicability and cannot cover all speed conditions;(2)the lack of consideration of the stiffness variations in the vehicle motion control model equations and their impact on the accuracy,computational efficiency,and numerical stability during the discretization of NMPC;(3)the increased burden of NMPC under extreme conditions and the need to take path tracking accuracy and dynamic stability of vehicle into consideration simultaneously.To address the above-mentioned problems,this study focuses on the research of NMPC vehicle motion control model,solution accuracy and computational efficiency,stability of vehicle motion control,and experimental verification.The main research contents are as follows:(1)Higher-level autonomous driving requires motion control algorithms to be applicable to a wider range of operating conditions.In the existing design of NMPC motion control algorithms,there is a problem of narrow applicability of a single vehicle motion control model.To address this issue,an Integrated Kinematic and Dynamic(IKD)model which combines vehicle kinematics and dynamics is proposed.In order to balance model accuracy and complexity,a coupling analysis of longitudinal and lateral dynamics is conducted for the proposed IKD model,and simplifications are made in the parts with weaker dynamic coupling.Experimental tests on the longitudinal and lateral motion characteristics are conducted using the Car Sim/Simulink co-simulation platform.A comparison between the IKD model,traditional kinematic model,and dynamic model verifies the good model accuracy of the proposed IKD model under both high and low-speed conditions.The constructed vehicle model provides the foundation for achieving NMPC-based autonomous driving motion control under all speed conditions.(2)The computational efficiency and control accuracy of NMPC motion control algorithms are crucial for their application in autonomous driving systems.Considering the impact of the stiffness of the vehicle model equations in the process of discretization of NMPC,traditional discrete methods require small discrete time steps to ensure solution accuracy and numerical stability when the equations exhibit stiffness.However,small discrete time steps can be inefficient.To address this issue,an Orthogonal Collocation on Finite Elements(OCFE)discretization-based NMPC motion control algorithm is proposed,which improves the computational efficiency of the NMPC solution process compared to traditional Euler discretization and fourth-order RungeKutta methods.When the stiffness of the vehicle model equations is weak,OCFE discretization achieves higher accuracy with the same discrete time step.The proposed OCFE discretizationbased NMPC motion control algorithm is compared to the traditional discrete method by Car Sim/Simulink co-simulation.The results demonstrate that the proposed control algorithm can address the computational inefficiency caused by the stiffness of the vehicle model equations and achieve higher control accuracy.(3)Under extreme conditions such as high-speed steering and low-friction road,the nonlinearity of the vehicle model intensifies,increasing the burden of online NMPC solution and affecting control real-time performance.Additionally,it becomes challenging to ensure vehicle dynamic stability under extreme conditions if trajectory tracking is considered without stability control.Loss of vehicle stability can result in the failure of trajectory tracking.To address this issue,trajectory tracking control algorithms and stability control algorithms are designed,taking into account path tracking accuracy,real-time performance,and vehicle stability under extreme conditions.To balance control accuracy and real-time performance of the trajectory tracking control algorithm under extreme conditions,a cascaded discretization method is proposed,leveraging the rolling optimization capability of NMPC.Furthermore,an effective NMPC prediction horizon extension strategy is proposed for specific high-speed driving scenarios.The stability control utilizes active rear-wheel steering to prevent excessive increase in the vehicle’s yaw angle,ensuring vehicle stability under extreme conditions.Real-time estimation of tire lateral forces is used to correct the lateral stiffness,enhancing the robustness of the control algorithm.The proposed motion control algorithm is validated through Car Sim/Simulink co-simulation,demonstrating that it can ensure vehicle stability under extreme conditions such as high-speed steering and low-adhesion road.Moreover,it achieves a good balance between path tracking accuracy and control real-time performance.(4)To further validate the effectiveness of the proposed control algorithms on a real vehicle,a downscaled intelligent car platform was built at a 1/10 scale.The precision of the car’s actuators was tested to ensure accurate execution of the specified commands.Based on this car platform,experiments were conducted and the results demonstrate the effectiveness of the proposed control algorithm.This paper focuses on the development of motion control methods based on NMPC.The research is conducted in various aspects,including the the vehicle motion control model,the precision and computational efficiency of the discretization of NMPC,vehicle motion control under extreme conditions,and experimental validation.Motion control algorithm that is applicable to a wide range of operating conditions,offers high control accuracy and computational efficiency,and ensures the safety and stability of vehicle operation is designed,which is expected to be applied in commercial autonomous driving systems.