摘要
针对目前档案馆机器人的避障路径较复杂、避障消耗时间较长等问题,研究结合机器学习中改进深度Q网络算法,搭建档案馆机器人自主避障算法。搭建档案馆机器人的空间运动模型和里程计算模型;对传统的深度Q网络算法进行优化,设计三重深度Q学习网络档案馆机器人避障算法。研究结果表明:所设计的算法具有较好的奖励曲线,当利用档案馆平面图搭建含有动态障碍点的实际动态障碍环境时,机器人自主避障寻优路径缩短到2.5 m,避障耗时仅8 s。
This research by combining the improved deep Q-network algorithm in machine learning,builds an autonomous obstacle avoidance algorithm for archive robot aiming at the complicated obstacle avoidance path and long consumed time of current archive robot.The spatial motion model and mileage calculation model of the archive robot are constructed,the traditional deep Q-network algorithm is optimized and the triple deep Q-learning network archive robot obstacle avoidance algorithm is designed.The results show that the designed algorithm has a good reward curve,and its autonomy obstacle avoidance seeking the optimal path is shortened to 2.5 m,and the obstacle avoidance time consumed only 8 s upon in a real dynamic obstacle environment containing dynamic obstacle points with an archive floor plan.
作者
顾昕
李雪
刘俊鹏
GU Xin;LI Xue;LIU Junpeng(Jilin Jiaohe Pumped Storage Co.,Ltd.,Jilin 132599,China)
出处
《机械制造与自动化》
2025年第4期248-253,共6页
Machine Building & Automation
关键词
机器人
自主避障
路径规划
机器学习
档案馆
robot
autonomous obstacle avoidance
path planning
machine learning
archives
