基于深度学习的智能驾驶车辆路径仿真
Intelligent driving vehicle routing simulation based on deep learning
投稿时间:2022-05-07  修订日期:2022-05-07
DOI:
中文关键词: 智能驾驶  深度强化学习  DDPG算法  经验回放机制  
英文关键词: Intelligent driving  Deep reinforcement learning  DDDPG algorithm  Experience playback mechanism  
基金项目:
作者单位邮编
何倩 陕西中医药大学 计算机实验中心 
仝武宁 陕西中医药大学 计算机实验中心 712000
摘要点击次数: 143
全文下载次数: 0
中文摘要:
      本文基于深度强化学习技术研究了智能驾驶中的路径规划问题,且在虚拟环境下进行模拟分析,对提出的路径规划算法性能做了验证研究。本文提出一种改进的经验回放机制ERDDPG (Experience Replay Deep Deterministic Policy Gradient )算法,对经验样本通过优先经验回放机制处理而设置不同的优先级,高质量的经验样本被优先采样,这种模式下网络的训练效率显著提高。在仿真实验中,本文的ERDDPG算法可完成智能驾驶的路径规划,学习效率较高,且智能车行驶的稳定性更好。
英文摘要:
      Based on the deep reinforcement learning technology, this paper studies the path planning problem in intelligent driving, simulates and analyzes it in the virtual environment, and verifies the performance of the proposed path planning algorithm. This paper proposes an improved experience replay mechanism erddpg (experience replay deep deterministic policy gradient) algorithm, which sets different priorities for the experience samples through the priority experience replay mechanism, and the high-quality experience samples are sampled first. In this mode, the training efficiency of the network is significantly improved. In the simulation experiment, the ERDDPG algorithm in this paper can complete the path planning of intelligent driving, with high learning efficiency and better driving stability of intelligent vehicle.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭