• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2026, Vol. 62 ›› Issue (8): 196-209.doi: 10.3901/JME.260278

• 特邀专辑:汽车线控底盘 • 上一篇    下一篇

扫码分享

基于PMSM无传感器控制的路面附着系数与轮胎侧偏特性参数估计

李浩然1, 周海超1, 王国林1, 张荣芸2, 赵春来3   

  1. 1. 江苏大学汽车与交通工程学院 镇江 212013;
    2. 安徽工程大学机械与汽车工程学院 芜湖 241000;
    3. 东风汽车集团有限公司研发总院 武汉 430000
  • 收稿日期:2025-06-03 修回日期:2025-12-10 出版日期:2026-04-20 发布日期:2026-06-12
  • 作者简介:李浩然,男,1998年出生,博士研究生。主要研究方向为智能轮胎设计与开发,车辆动力学及其控制。E-mail: terteye@foxmail.com;周海超(通信作者),男,1984年出生,博士,教授,博士研究生导师。主要研究方向为汽车轮胎结构设计与性能优化、智能轮胎设计与开发和车辆动力学及其控制。E-mail: hczhou@ujs.edu.cn
  • 基金资助:
    国家自然科学基金(52272366,52072156);中国博士后科学基金(2020M682269)资助项目。

Road Friction Coefficient and Tire Cornering Characteristics Parameters Estimation Based on PMSM Sensorless Control

LI Haoran1, ZHOU Haichao1, WANG Guolin1, ZHANG Rongyun2, ZHAO Chunlai3   

  1. 1. School of Automobile and Traffic Engineering, Jiangsu University, Zhenjiang 212013;
    2. School of Mechanical and Automotive Engineering, Anhui Polytechnic University, Wuhu 241000;
    3. Dongfeng Motor Corporation Research & Development Institute, Wuhan 430000
  • Received:2025-06-03 Revised:2025-12-10 Online:2026-04-20 Published:2026-06-12

摘要: 针对分布式驱动电动汽车(Distributed drive electric vehicle,DDEV)路面附着系数与轮胎侧偏特性参数估计精度不足的问题,结合多传感器融合与物理约束神经网络建模理论对路面附着系数与轮胎侧偏特性参数进行估计。首先,基于最大相关熵(Maximum correntropy,MC)准则与平方根正交容积卡尔曼滤波(Square-root cubature quadrature Kalman filter,SCQKF)推导得到MC-SCQKF算法,该算法通过优化量测误差协方差矩阵和正交容积点采样来提升非高斯噪声干扰下对永磁同步电机(Permanent magnet synchronous motor,PMSM)转速与转子位置的估计精度,同时增强系统的鲁棒性和状态估计的收敛性。然后,利用MC-SCQKF实时估计横摆角速度、质心侧偏角、纵向速度和路面附着系数,并将其作为特征来构建预测轮胎侧向力的物理信息神经网络(Tire physics-informed neural networks,TirePINN)模型。最后,根据车辆状态参数估计值计算的轮胎侧偏角和侧向力的预测值,对前后轮侧偏刚度进行拟合,形成面向轮胎侧偏特性的自验证闭环观测系统。硬件在环试验表明,拟合的前轮侧偏刚度的误差为1.49%,后轮侧偏刚度误差为1.37%。

关键词: 分布式驱动电动汽车, 平方根正交容积卡尔曼滤波, 物理信息神经网络, 路面附着系数, 轮胎侧偏特性参数

Abstract: Aiming at the insufficient estimation accuracy of road adhesion coefficient and tire cornering characteristics parameters for distributed drive electric vehicle(DDEV), the road adhesion coefficient and tire cornering characteristics parameters estimation are constructed by combining multi-sensor fusion and physical constraint neural network modeling theories. First, the maximum correntropy(MC) criterion is integrated with square-root cubature quadrature Kalman filter(SCQKF) to construct the MC-SCQKF algorithm, where the measured covariance matrix is optimized and quadrature cubature sampling is employed to achieve accurate estimation of PMSM speed and rotor position under non-Gaussian noise, simultaneously enhancing the robustness of the system and the convergence of state estimation. Second, enabling real-time estimation of yaw rate, sideslip angle, longitudinal speed, and road friction coefficient through MC-SCQKF, and they are utilized as features to develop the physics-informed neural network for tire lateral force prediction(TirePINN) model. Finally, the cornering stiffness of front and rear tires is fitted using predicted lateral forces and calculated slip angles by the estimated vehicle state parameters, forming a self-validating closed-loop observation system for tire cornering characteristics. Hardware-in-the-loop tests show fitted errors of 1.49%(front) and 1.37%(rear) for tire cornering stiffness.

Key words: distributed drive electric vehicle, square-root cubature quadrature Kalman filter, physics-informed neural network, road friction coefficient, tire cornering characteristic parameters

中图分类号: