Scholarlink

Autonomous driving has the ability to reduce the frequency of traffic accidents by eliminating the possibility of human error, which is a major cause of accidents. Considering various scenarios in autonomous driving, crossing intersections is one of the most complex scenarios. This study addresses a solution for autonomous vehicles (AVs) based on deep reinforcement learning (DRL), which will be helpful for crossing intersections safely and efficiently. DRL-enabled decision-making framework is used in this research to train the AVs to drive through intersections without any collisions. The interaction between autonomous vehicles (AVs) and other vehicles was modeled using the Markov decision process (MDP), and the optimal driving policy was obtained using the deep Q-network algorithm. Performance was measured based on the success rate and missed opportunities.

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

Deep Reinforcement Learning Approach for Autonomous Vehicles to Cross Junctions

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Samsani Surekha,
  • R. V. S. Lalitha,
  • Munnangi Srikanya,
  • Sureddy Jaya Sri,
  • Radha Kinjagi,
  • Dande Keerthana

摘要

Autonomous driving has the ability to reduce the frequency of traffic accidents by eliminating the possibility of human error, which is a major cause of accidents. Considering various scenarios in autonomous driving, crossing intersections is one of the most complex scenarios. This study addresses a solution for autonomous vehicles (AVs) based on deep reinforcement learning (DRL), which will be helpful for crossing intersections safely and efficiently. DRL-enabled decision-making framework is used in this research to train the AVs to drive through intersections without any collisions. The interaction between autonomous vehicles (AVs) and other vehicles was modeled using the Markov decision process (MDP), and the optimal driving policy was obtained using the deep Q-network algorithm. Performance was measured based on the success rate and missed opportunities.