基于 TD3-PER 的主动悬架多目标协同控制

doi:10.16638/j.cnki.1671-7988.2026.018.007

汽车实用技术 ›› 2026, Vol. 51 ›› Issue (18): 34-42.DOI: 10.16638/j.cnki.1671-7988.2026.018.007

• 设计研究 • 上一篇

基于 TD3-PER 的主动悬架多目标协同控制

邓彭宇 1,2，卢荡 2，高雄 2

1.福建理工大学电子电气与物理学院； 2.福建理工大学新能源车辆运动控制研究院

发布日期:2026-09-18
通讯作者: 邓彭宇
作者简介:邓彭宇（2000－），女，硕士，研究方向为深度强化学习、主动悬架控制、后轮主动转向控制与分布式驱动控制
基金资助:
福建省财政厅项目（GY-Z23269）；福建省/市科技重大项目（2024-ZD-011）

Multi-objective cooperative control of active suspension systems based on TD3-PER

DENG Pengyu 1,2 , LU Dang2 , GAO Xiong2

1.College of Electronic, Electrical and Physics, Fujian University of Technology; 2.New Energy Vehicle Motion Control Research Institute, Fujian University of Technology

Published:2026-09-18
Contact: DENG Pengyu

摘要/Abstract

摘要： 针对传统深度强化学习（DRL）在车辆主动悬架控制中，样本利用效率低和多目标优化效果不足的问题，提出一种融合优先级经验回放（PER）机制的双延迟深度确定性策略梯度（TD3）算法（TD3-PER）。该算法通过优先筛选高价值样本，显著提高样本利用率，加速高维状态空间中的策略收敛。同时，设计了一种综合垂向加速度与俯仰角加速度的加权目标函数自适应调节，实现多自由度振动的协同优化。仿真实验表明，在 C 级随机路面下，相较于被动悬架，TD3-PER 算法将车身垂向加速度均方根值降低 38.79%，俯仰角加速度均方根值降低 45.24%。在 20～120 km/h 车速范围内，车身加速度优化率达 36.26%～52.46%，俯仰角加速度优化率达 20.99%～48.92%，且性能随路面平滑度提升而增强，具有更好的高效性、鲁棒性及多目标优化能力。

关键词: 主动悬架；深度强化学习；优先经验回放；多目标优化

Abstract: To address the issues of low sample efficiency and insufficient multi-objective optimization in traditional deep reinforcement learning (DRL) for vehicle active suspension control, a twin delayed deep deterministic policy gradient (TD3) algorithm integrated with a prioritized experience replay (PER) mechanism (TD3-PER) is proposed. This algorithm significantly improves sample efficiency and accelerates policy convergence in high-dimensional state spaces by prioritizing high-value samples. Additionally, an adaptive weighted objective function combining vertical acceleration and pitch acceleration is designed to achieve coordinated optimization of multi-degree-of-freedom vibrations. Simulation experiments demonstrate that, under class C random road conditions, the TD3-PER algorithm reduces the root mean square (RMS) of vertical body acceleration by 38.79% and the RMS of pitch angular acceleration by 45.24% compared to passive suspension. Within a vehicle speed range of 20～120 km/h, the optimization rate for body acceleration ranges from 36.26% to 52.46%, and for pitch angular acceleration from 20.99% to 48.92%, with performance improving as road smoothness increases. The algorithm exhibits superior efficiency, robustness, and multi-objective optimization capabilities.

Key words: active suspension; deep reinforcement learning; prioritized experience replay; multiobjective optimization

邓彭宇 1,2，卢荡 2，高雄 2. 基于 TD3-PER 的主动悬架多目标协同控制[J]. 汽车实用技术, 2026, 51(18): 34-42.

DENG Pengyu 1,2 , LU Dang2 , GAO Xiong2. Multi-objective cooperative control of active suspension systems based on TD3-PER[J]. Automobile Applied Technology, 2026, 51(18): 34-42.

参考文献

[ 1 ] Du Guodong,Zou Yuan,Zhang Xudong,et al.Hierarchical motion planning and tracking for autonomous vehicles using global heuristic based potential field and reinforcement learning based predictive control[J]. IEEE Transactions on Intelligent Transportation Systems,2023,24(8):8304-8323. [ 2 ] Garikapati D,Shetiya S S.Autonomous vehicles:evolution of artificial intelligence and the current industry landscape[J].Big Data and Cognitive Computing,2024, 8(2):42. [ 3 ] Karnopp D C,Croby M J,Harwood R A.Vibration control using semi active force generators[J].Journal of Engineering for Industry,1974,96(2):619-626. [ 4 ] Valá?ek M,Novák M,?ika Z,et al.Extended ground hook new concept of semi active control of truck's suspension[J].Vehicle System Dynamics,1996,27(5/6): 289-303. [ 5 ] Ahmadian M,Vahdati N.Transient dynamics of semiactive suspensions with hybrid control[J].Journal of Intelligent Material Systems and Structures,2006,17(2):145-153. [ 6 ] Kumar V,Rana K P S.A novel fuzzy PID controller for nonlinear active suspension system with an electro hydraulic actuator[J].Journal of the Brazilian Society of Mechanical Sciences and Engineering,2023,45:189. [ 7 ] 熊新,陈佳玲,魏金呈.主动悬架的遗传算法优化 LQR 控制策略研究[J].盐城工学院学报(自然科学版), 2024,37(1):60-66. [ 8 ] 于曰伟,周长城,赵雷雷,等.基于交替迭代的车辆主动悬架 LQG 控制器设计[J].山东大学学报(工学版), 2017,47(4):50-56. [ 9 ] 冯桂珍,赵东鹏,李韶华.基于分数阶的空气弹簧建模及电动汽车主动悬架控制研究[J].汽车工程,2024,46 (7):1282-1293,1301. [10] 胡耀中,刘俊.基于遗传算法优化的汽车主动悬架 LQR 控制研究[J].农业装备与车辆工程,2017(7):45- 48. [11] Li Mei,Xu Jie,Wang Zelong,et al.Optimization of the semi active suspension control of BP neural network PID based on the sparrow search algorithm[J].Sensors, 2024,24(6):1757. [12] Yong H,Seo J,Kim J,et al.Suspension control strategies using switched soft actor critic models for real roads [J].IEEE Transactions on Industrial Electronics,2023, 70(1):824-832. [13] Wang Yang,Wang Cheng,Zhao Shijie,et al.Research on deep reinforcement learning control algorithm for active suspension considering uncertain time delay[J]. Sensors,2023,23(18):7827. [14] Wang Cheng,Cui Xiaoxian,Zhao Shijie,et al.Enhancing vehicle ride comfort through deep reinforcement learning with expert guided soft hard constraints and system characteristic considerations[J].Advanced Engineering Informatics,2024,59:102328. [15] Wang Cheng,Cui Xiaoxian,Zhao Shijie,et al.A deep reinforcement learning based active suspension control algorithm considering deterministic experience tracing for autonomous vehicle[J].Applied Soft Computing, 2024,153:111259. [16] Deng Mingxing,Sun Dongxu,Zhan Liu,et al.Advancing active suspension control with TD3-PSC:integrating physical safety constraints into deep reinforcement learning[J].IEEE Access,2024,12:115628-115641. [17] ISO 8608:2016 Mechanical vibration road surface profiles-reporting of measured data[S]. [18] ISO 2631-1:1997 Mechanical vibration and shock evaluation of human exposure to whole body vibrationPart 1:General requirements[S].

基于 TD3-PER 的主动悬架多目标协同控制

Multi-objective cooperative control of active suspension systems based on TD3-PER

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

编辑推荐

Metrics