Scholarlink

<p>This study introduces an electric multi-axle truck (EMAT) featuring a new power system, along with collaborative active steering control (ASC) and direct yaw moment control (DYC) to tackle steering problems. The EMAT design includes a drive motor and differential on the middle axle, as well as two independent motors on the rear axle. This rear-wheel distributed drive system, supported by the drive motor, meets heavy-load demands, reduces rear motor stress and facilitates collaborative ASC and DYC. The EMAT also utilizes the broad critic deep actor (BCDA) reinforcement learning (RL) algorithm to train ASC and DYC, which demands fewer computational resources and training time than traditional RL controllers due to its optimized training process. Prototype tests validate the effectiveness of the EMAT and RL-based BCDA controller in terms of significant improvement on steering performance (including over 80% reduction in training time compared to the deep deterministic policy gradient (DDPG) algorithm). The proposed EMAT and RL-based BCDA controller show promise in truck design and dynamic control.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Design and Experimental Evaluation of a Novel Multi-axle Electric Truck with Broad-Deep Learning-Based Collaborative Controllers for Active Steering and Direct Yaw Moment Control

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Pak Kin Wong,
  • Xiaozheng Wang,
  • Shiron Thalagala

摘要

This study introduces an electric multi-axle truck (EMAT) featuring a new power system, along with collaborative active steering control (ASC) and direct yaw moment control (DYC) to tackle steering problems. The EMAT design includes a drive motor and differential on the middle axle, as well as two independent motors on the rear axle. This rear-wheel distributed drive system, supported by the drive motor, meets heavy-load demands, reduces rear motor stress and facilitates collaborative ASC and DYC. The EMAT also utilizes the broad critic deep actor (BCDA) reinforcement learning (RL) algorithm to train ASC and DYC, which demands fewer computational resources and training time than traditional RL controllers due to its optimized training process. Prototype tests validate the effectiveness of the EMAT and RL-based BCDA controller in terms of significant improvement on steering performance (including over 80% reduction in training time compared to the deep deterministic policy gradient (DDPG) algorithm). The proposed EMAT and RL-based BCDA controller show promise in truck design and dynamic control.