汽车工程 ›› 2024, Vol. 46 ›› Issue (11): 2046-2058.doi: 10.19562/j.chinasae.qcgc.2024.11.011

• • 上一篇    下一篇

考虑不同时域的商用车预见性巡航控制

耿小虎1,付尧1(),王杰1,雷雨龙1,刘卫东1,王玉海2,刘科1   

  1. 1.吉林大学,汽车仿真与控制国家重点实验室,长春 130000
    2.一汽解放青岛汽车有限公司,青岛 266000
  • 收稿日期:2024-06-02 修回日期:2024-07-09 出版日期:2024-11-25 发布日期:2024-11-22
  • 通讯作者: 付尧 E-mail:fu_yao@jlu.edu.cn
  • 基金资助:
    四川省重点研发项目(2023YFG0068);青岛市科技计划项目(22-5-1-yfzt-4-jch);山东省泰山产业领军人才工程项目(tscx202211119)

Predictive Cruise Control for Commercial Vehicles Considering Different Time Domains

Xiaohu Geng1,Yao Fu1(),Jie Wang1,Yulong Lei1,Weidong Liu1,Yuhai Wang2,Ke Liu1   

  1. 1.Jilin University,State Key Laboratory of Automotive Simulation and Control,Changchun 130000
    2.FAW Jiefang Qingdao Automobile Co. ,Ltd. ,Qingdao 266000
  • Received:2024-06-02 Revised:2024-07-09 Online:2024-11-25 Published:2024-11-22
  • Contact: Yao Fu E-mail:fu_yao@jlu.edu.cn

摘要:

预见性巡航控制(predictive cruise control,PCC)在规划层以预测节能为目标进行长时域的速度规划,执行层对规划速度进行短时域的跟踪控制。由于规划层与执行层有着不同时间尺度步长要求,在系统设计中很难将二者置于一个优化控制问题中。因此,本文采用分层控制思想,在规划层基于改进的双延迟深度确定性策略梯度算法(twin delayed deep deterministic policy gradient algorithm,TD3)获得预测时域内长周期的规划速度;在执行层基于模型预测控制(model predictive control,MPC)以规划速度为参考速度,同时考虑发动机油耗特性和变速器换挡规律,对规划速度在短时域内作进一步的经济性优化,并进行跟踪控制。硬件在环验证结果表明,将改进的TD3与MPC相结合可以改善PCC在规划与执行中的时间尺度不一致问题,并有效降低重型商用车在巡航过程中的燃油消耗量和换挡频次。

关键词: 预见性巡航, 速度规划与控制, 深度强化学习, 模型预测控制

Abstract:

Predictive cruise control (PCC) performs long-term speed planning at the planning layer with the objective of predicting energy savings and short-term tracking control for the vehicle speed at the execution layer. Integrating these layers into a single optimal control problem poses significant challenges in system design due to the different time scale step requirements between the planning layer and the execution layer. To address this challenge, a hierarchical control approach is adopted in this paper. At the planning layer, an improved twin delayed deep deterministic policy gradient (TD3) algorithm is utilized to determine the long-term planning speed over the prediction horizon. Meanwhile, at the execution layer, based on model predictive control (MPC), taking the planned vehicle speed as the reference speed and considering engine fuel consumption characteristics and transmission shift laws, further economic optimization and tracking control of the planned speed are carried out in the short term. The hardware-in-the-loop (HIL) validation results show that combining the improved TD3 algorithm with MPC effectively resolves the time scale inconsistency between planning and execution in PCC, which can significantly reduce both fuel consumption and shift frequency during the cruising of heavy-duty commercial vehicles.

Key words: predictive cruising, speed planning and control, deep reinforcement learning, model predictive control