主办：陕西省汽车工程学会
ISSN 1671-7988  CN 61-1394/TH
创刊：1976年
汽车实用技术

汽车实用技术 ›› 2025, Vol. 50 ›› Issue (21): 11-20.DOI: 10.16638/j.cnki.1671-7988.2025.021.003

• 智能网联汽车 • 上一篇    

基于 DE_PSO 优化的无人驾驶车辆 LQR 横向控制策略研究

骆远鹏，张成涛*，赵晓卓，黎俊宏   

  1. 广西科技大学 机械与汽车工程学院
  • 发布日期:2025-11-04
  • 通讯作者: 张成涛
  • 作者简介:骆远鹏（1999－），男，硕士研究生，研究方向为自动驾驶规划及控制； 通信作者：张成涛（1978－），男，博士，副教授，研究方向为车辆智能化控制技术
  • 基金资助:
    广西科技大学硕士研究生创新项目（GKYC202410）

Research on LQR Lateral Control Strategy for Driverless Vehicles Based on DE_PSO Optimization

LUO Yuanpeng, ZHANG Chengtao* , ZHAO Xiaozhuo, LI Junhong   

  1. School of Mechanical and Automotive Engineering, Guangxi University of Science and Technology
  • Published:2025-11-04
  • Contact: ZHANG Chengtao

PDF

摘要： 针对传统线性二次型调节器（LQR）在无人驾驶车辆横向控制中存在权重矩阵依赖人 工经验、难以适应复杂工况的问题，文章提出一种融合差分进化（DE）算法与粒子群优化（PSO） 算法的混合优化方法（简称 DE_PSO）对 LQR 权重矩阵进行整定。基于二自由度车辆动力学 模型建立横向跟踪误差状态空间模型，并引入前馈补偿以提高稳态性能。然后，在 PSO 算法 中周期性引入差分进化操作，并结合非线性递减惯性权重和精英保留策略，有效提升了算法 的全局搜索能力与收敛稳定性，从而实现了对 LQR 权值的自适应整定。CarSim/Simulink 联 合仿真结果表明，与基于 PSO 的 LQR 方法相比，所提方法的最大横向误差和航向误差分别 降低了 50.7%和 37.5%，平均计算时间减少了 54.2%，在不同速度工况下均表现出稳定的轨迹 跟踪精度和良好的鲁棒性。

关键词: 横向跟踪控制；线性二次型调节器；差分进化；粒子群优化；CarSim/Simulink

Abstract: In order to solve the problem of reliance on manual tuning and the poor adaptability of traditional linear quadratic regulator (LQR) controllers in autonomous vehicle lateral control, this paper proposes a hybrid optimization method that integrates differential evolution (DE) and particle swarm optimization (PSO), referred to as DE_PSO, for tuning the LQR weight matrices. Firstly, a lateral tracking error state-space model is constructed based on a two-degree-of-freedom vehicle dynamics model, and feedforward compensation is introduced to enhance steady-state performance.Then, DE operations are periodically incorporated into the PSO framework, together with a nonlinear inertia weight decay and an elitism retention mechanism, to enhance global search performance and convergence stability, thereby facilitating adaptive tuning of LQR weight matrices. Finally, The results of CarSim/Simulink joint simulation show that compared with the PSO based LQR method, the proposed method reduces the maximum lateral error and heading error by 50.7% and 37.5% respectively, reduces the average calculation time by 54.2%, and shows stable trajectory tracking accuracy and good robustness under different speed conditions.

Key words: lateral tracking control; linear quadratic regulator; differential evolution; particle swarm optimization: CarSim/Simulink