摘要
针对隧道巡检机器人的应用需求,普通RGB相机只能得到二维图像信息,无法有效获取深度信息以及被遮挡的障碍物位置信息。因此无法有效引导机械臂避开障碍物。通过图像和机械臂末端反馈力构建综合环境感知系统,设计奖惩函数作为无模型强化学习算法的环境信息,引导机械臂避开障碍达到目标点。最后实验验证了此方法的有效性。
According to the application requirements of the tunnel inspection robots,ordinary RGB cameras can only get 2 d images,and cannot effectively obtain depth information and position of the blocked obstacles.Therefore,the manipulator cannot be effectively guided to avoid obstacles.The integrated environmental perception system is constructed by means of image and feedback force at the end of the manipulator.The reward and punishment function is designed as the environment information of model-free reinforcement learning algorithm.It directs the arm to avoid obstacles and reach the target point.Finally,experiments verify the validity of this method.
作者
许继葵
石银霞
单鲁平
徐研
蒋勋
XU Ji-kui;SHI Yin-xia;SHAN Lu-ping;XU Yan;JIANG Xun(Guangzhou Power Supply Co.,Ltd.,Guangzhou 510000 China;PertroChina Southwest Oil&Gastield Company,Chengdu 610051 China)
出处
《自动化技术与应用》
2020年第1期88-92,共5页
Techniques of Automation and Applications
基金
运维关键技术研究与工程实践（编号GZHKJXM20160041）
关键词
巡检机器人
避障
强化学习
机器人
inspection robot
obstacle avoidance
reinforcement learning
Robot
