基于双决斗深度Q网络的自动换道决策模型  

Automatic Lane Change Decision Model Based on Dueling Double Deep Q-network

作　　者：张雪峰[1] 王照乙 ZHANG Xue-feng;WANG Zhao-yi(School of Sciences,Northeastern University,Shenyang 110819,China)

机构地区：[1]东北大学理学院,辽宁沈阳110819

出　　处：《东北大学学报（自然科学版）》2023年第10期1369-1376,共8页Journal of Northeastern University(Natural Science)

基　　金：国家重点研发计划项目(2020YFB1710003).

摘　　要：汽车自动变道需要在保证不发生碰撞的情况下,以尽可能快的速度行驶,规则性地控制不仅对意外情况不具有鲁棒性,而且不能对间隔车道的情况做出反应.针对这些问题,提出了一种基于双决斗深度Q网络(dueling double deep Q-network,D3QN)强化学习模型的自动换道决策模型,该算法对车联网反馈的环境车信息处理之后,通过策略得到动作,执行动作后根据奖励函数对神经网络进行训练,最后通过训练的网络以及强化学习来实现自动换道策略.利用Python搭建的三车道环境以及车辆仿真软件CarMaker进行仿真实验,得到了很好的控制效果,结果验证了本文算法的可行性和有效性.Automatic lane change of vehicles requires driving at the fastest possible speed while ensuring no collision situations.However,regular control is not robust enough to handle unexpected situations or respond to lane separation.To solve these problems,an automatic lane change decision model based on dueling double deep Q-network(D3QN)reinforcement learning model is proposed.The algorithm processes the environmental vehicle information fed back by the internet of vehicles,and then obtains actions through strategies.After the actions are executed,the neural network is trained according to given reward function,and finally the automatic lane change strategy is realized through the trained network and reinforcement learning.The three-lane environment built by Python and the vehicle simulation software CarMaker are used to carry out simulation experiments.The results show that the algorithm proposed has a good control effect,making it feasible and effective.

关 键 词：车道变换 自动驾驶 强化学习 深度学习 深度强化学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]