Scholarlink

This study focuses on autonomous driving technology and proposes an optimal path design method to reduce the risk of traffic accidents. The autonomous driving system is divided into three layers: “perception,” “decision-making,” and “control,” each layer requires a distinct task. In previous studies, control methods using Model Predictive Control (MPC) have been proposed for the control layer of the system design. On the one hand, this study aims to introduce deep reinforcement learning into the decision-making layer of the system design. The objective of this study is to propose a path generation method using deep reinforcement learning for the MPC method applied to the automatic driving system. The effectiveness of the proposed method is verified by numerical simulation.

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

Path Generation Using Deep Reinforcement Learning for Model Predictive Control of Automatic Driving

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Tassei Uematsu,
  • Tomoaki Hashimoto

摘要

This study focuses on autonomous driving technology and proposes an optimal path design method to reduce the risk of traffic accidents. The autonomous driving system is divided into three layers: “perception,” “decision-making,” and “control,” each layer requires a distinct task. In previous studies, control methods using Model Predictive Control (MPC) have been proposed for the control layer of the system design. On the one hand, this study aims to introduce deep reinforcement learning into the decision-making layer of the system design. The objective of this study is to propose a path generation method using deep reinforcement learning for the MPC method applied to the automatic driving system. The effectiveness of the proposed method is verified by numerical simulation.