Scholarlink

<p>Economic driving technology is crucial for the lifecycle of commercial vehicles. Among optimization-based control algorithms, Pontryagin’s Maximum Principle (PMP) offers strong adaptability and is advantageous for embedded deployment. However, fuel economy depends on many factors, and PMP’s nonlinear equations hinder analytical solutions. This study proposes a PMP-based Physics-Informed Neural Network (PMP-PINN) approach, embedding physical constraints into the loss function to approximate PMP differential equation solutions and generate fuel-optimal speed profiles. The method features a compact structure and high computational efficiency, enabling real-time embedded applications. Simulations show the PMP-PINN achieves &lt; 0.5% deviation from Dynamic Programming (DP) while cutting computation time nearly 100-fold, outperforming conventional algorithms. This provides an effective solution and practical reference for vehicle dynamics control and onboard energy-efficient driving.</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

Energy-Saving Cruise Control for Commercial Vehicles Based on Physics-Informed Neural Networks (PINN)

  • 期刊论文

  • 发表日期：

    2026年2月26日

查看全文
  • Xingkun Li,
  • Guohui Wang,
  • Bo Gao,
  • Guangyu Tian,
  • Ziwang Lu,
  • Yuhai Wang

摘要

Economic driving technology is crucial for the lifecycle of commercial vehicles. Among optimization-based control algorithms, Pontryagin’s Maximum Principle (PMP) offers strong adaptability and is advantageous for embedded deployment. However, fuel economy depends on many factors, and PMP’s nonlinear equations hinder analytical solutions. This study proposes a PMP-based Physics-Informed Neural Network (PMP-PINN) approach, embedding physical constraints into the loss function to approximate PMP differential equation solutions and generate fuel-optimal speed profiles. The method features a compact structure and high computational efficiency, enabling real-time embedded applications. Simulations show the PMP-PINN achieves < 0.5% deviation from Dynamic Programming (DP) while cutting computation time nearly 100-fold, outperforming conventional algorithms. This provides an effective solution and practical reference for vehicle dynamics control and onboard energy-efficient driving.