基于机器学习的自动巡检系统优化设计及性能测试研究  

Research on the optimization design and performance test of automatic inspection system based on machine learning

作　　者：王坤 甄超 罗沙 WANG Kun;ZHEN Chao;LUO Sha(State Grid Anhui Electric Power Co.,Ltd.,Hefei 230061,China)

机构地区：[1]国网安徽省电力有限公司,安徽合肥230061

出　　处：《粘接》2024年第5期141-144,共4页Adhesion

基　　金：国家发展和改革委项目(项目编号:Zsf20220007649326)。

摘　　要：为了提高巡检效率和避障精度,提出了基于机器学习的机器人自动巡检系统。根据机器人巡检电机驱动器的内部结构,给出电机驱动的技术参数,利用巡检电机驱动器的外部接口,将巡检电机驱动器与其他硬件连接,通过引入一个调理电路,调节电机驱动器的输出电压,设计了机器人自动巡检电路;软件设计中,规划了机器人巡检路径,利用机器学习算法推算出机器人的巡检轨迹,计算出机器人巡检过程中方位的置信度,结合机器人自动巡检算法的设计,实现了机器人的自动巡检。实验结果表明,所提系统在机器人自动巡检中,可以将巡检效率和避障精度分别控制在83%和85%以上。In order to improve patrol inspection efficiency and obstacle avoidance accuracy,the development and application research of robot automatic patrol system technology based on machine learning was proposed.According to the internal structure of the robot patrol motor driver,the technical parameters of the motor driver were given,and the patrol motor driver was connected with other hardware by using the external interface of the patrol motor driver.By introducing a conditioning circuit to adjust the output voltage of the motor driver,the robot automatic patrol circuit was designed.In the software design,based on the path planning goal of robot automatic inspection,the robot inspection path was planned,the robot inspection trajectory was calculated by using machine learning algorithm,and the confidence of the robot's orientation in the inspection process was calculated.Combined with the design of robot automatic inspection algorithm,the robot automatic inspection was realized.The experimental results showed that the system in this paper could control the inspection efficiency and obstacle avoidance accuracy above 83%and 85%respectively in robot automatic inspection.

关 键 词：机器学习 自动巡检 电机驱动器 机器人 路径规划 避障精度 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置] TM307[自动化与计算机技术—控制科学与工程]