基于深度学习的巡检机器人避障轨迹自动控制系统设计  

Design of Automatic Control System for Obstacle Avoidance Trajectory ofInspection Robot Based on Deep Learning

作　　者：乔道迹[1] QIAO Daoji(School of computer science and technology,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学计算机科学与技术学院,太原030051

出　　处：《计算机测量与控制》2024年第5期129-136,共8页Computer Measurement &Control

摘　　要：为了提高巡检机器人在复杂环境下的避障能力,使机器人能够安全地完成巡检任务,设计基于深度学习的巡检机器人避障轨迹自动控制系统;设计由CCD传感器、信号处理芯片等设备组成的工业智能视觉CCD相机,基于FPGA和USB2.0的视频采集卡传输采集数据,完成硬件部分的设计;在软件设计中,对采集的图像实施目标分割、双目目标匹配等预处理,通过对摄像头实施双目视觉标定获取障碍物空间位置三维信息,基于深度学习中的CRNN设计机器人自主避障规划网络模型,并设计模糊轨迹控制器,实现避障中的轨迹自动控制;系统测试结果表明,设计系统最终成功避开了3个动态障碍物,最大轨迹控制误差的最大值为1.45°,最小轨迹控制误差的最大值为0.62°,动态避障巡检速度始终在3.5 m/s左右,表现出了精准而稳定的轨迹控制效果。In order to improve the obstacle avoidance capability of inspection robots in complex environments and safely complete inspection tasks,an inspection robot obstacle avoidance trajectory automatic control system based on deep learning is designed.In the hardware design,the industrial intelligent vision CCD camera consists of CCD sensors,signal processing chips,the collected data is transmitted through the video acquisition card of FPGA and USB2.0.In the software design,the preprocessing techniques such as target segmentation and binocular object matching are applied to the captured images.The binocular visual calibration on the cameras is performed to obtain the three-dimensional spatial information of obstacles.Furthermore,an autonomous obstacle avoidance planning neural network model based on convolutional recurrent neural network(CRNN)is developed,and a fuzzy trajectory controller is designed to achieve automatic control of the trajectory during obstacle avoidance.The test results show that the system successfully avoids three dynamic obstacles,with a maximum trajectory control error of 1.45°and minimum trajectory control error of 0.62°.The dynamic obstacle avoidance inspection speed is always maintained at around 3.5 m/s,exhibiting the precise and stable trajectory control performance.

关 键 词：深度学习 视频采集卡 双目视觉标定 巡检机器人避障 轨迹控制 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]