基于随机森林的非线性车辆模型构建与高精度轨迹跟踪控制研究  

Research on Nonlinear Vehicle Model Construction and High-Precision Trajectory Tracking Control Based on Random Forest

作　　者：龙尧成 孙文 张忠[3] 何永 刘桂均 LONG Yaocheng;SUN Wen;ZHANG Zhong;HE Yong;LIU Guijun(College of Automotive Engineering,Changzhou Institute of Technology,Changzhou 213032,Jiangsu,China;National Key Laboratory of Automotive Chassis Integration and Bionics,Jilin University,Changchun 130025,China;CARTARC(Tianjin)Automotive Engineering Research Institute Co.,Ltd.,Tianjin 300300,China)

机构地区：[1]常州工学院汽车工程学院,江苏常州213032 [2]吉林大学汽车底盘集成与仿生全国重点实验室,长春130025 [3]中汽研(天津)汽车工程研究院有限公司,天津300300

出　　处：《汽车工程学报》2025年第2期197-210,共14页Chinese Journal of Automotive Engineering

基　　金：国家自然科学基金项目(62273061,52272365);吉林省中青年科技创新创业卓越人才(团队)项目(创新类)(20230508050RC);吉林省自然科学基金项目(20220101200JC);江苏省高等学校基础科学(自然科学)研究重大项目(22KJA580001,21KJA130001);江苏省高等学校基础科学(自然科学)研究面上项目(22KJB580001);常州市科技计划项目(CZ2022030)。

摘　　要：针对传统控制方案描述的车辆动力学特性精度有限,对预期状态的高精确跟踪难以实现,介绍了一种数据驱动的模型预测路径跟踪控制方法。基于随机森林方法构建了车辆状态参数观测器,并通过该观测器剖析了车辆动力学非线性映射关系以优化控制器的底层数学模型,以减少外部环境和车辆自身机械结构扰动对控制性能的不利影响。根据模型预测控制机理,结合车辆动力学映射关系构建整车状态空间方程,分析了车辆状态在局部范围内的线性变化规律,并以最优四轮附着利用率为目标,设计计算最优方向盘转角与四轮驱动力的二次规划代价函数。仿真结果表明,所提出的控制方案能在扰动存在的情况下避免过大的车身状态波动,并且在无扰动的行驶路段中也能保持较低的轮胎附着利用率,实现安全、稳定的高精度跟踪。In view of the limited accuracy of vehicle dynamics described by traditional control schemes,it is difficult to achieve high precision tracking of the expected state.Therefore,a data-driven model predictive control method for path tracking is introduced.Firstly,a vehicle state parameter observer was constructed using the random forest method.Based on this observer,the nonlinear mapping relationship of vehicle dynamics was analyzed to optimize the controller's underlying mathematical model,thereby reducing the adverse effects of external environmental and mechanical structural disturbances on control performance.Secondly,according to the model predictive control mechanism and vehicle dynamics mapping relationship,the vehicle state space equation was constructed.The linear pattern of vehicle state changes within the local range was analyzed.The quadratic programming cost function for optimizing the steering wheel angle and four-wheel driving force was designed and calculated,aiming to achieve the optimal utilization rate of fourwheel adhesion.Finally,the simulation results show that the proposed control scheme can prevent excessive fluctuations in vehicle body state in the presence of disturbances,and it also maintains a low utilization rate of tire adhesion on undisturbed road sections,achieving safe,stable and high-precision tracking.

关 键 词：机器学习 非线性映射 高精度跟踪控制 模型预测控制 自动驾驶 

分 类 号：U463.5[机械工程—车辆工程]