Research On Key Technologies Of Human-Machine Cooperation For Dual-Arm Operation Robots In Distribution Network

The distribution network is an important component of the power system,and live working can ensure uninterrupted power supply to users,which has significant social and economic benefits.However,the common overhead circuit spacing of 10kV distribution networks is small,and frontline personnel working under harsh geographical or weather conditions for live maintenance have long working hours,high difficulty in operation,high labor intensity,and high personal safety risks.The use of robots to replace manual work for live-line maintenance and repair has become an important development trend and research hotspot.However,the current distribution network live working robots generally have prominent problems such as poor flexibility,weak perception ability,low work efficiency,and poor environmental adaptability,making it difficult to adapt to the complex and diverse needs of live working.This article mainly focuses on the key technologies of human-machine collaboration for dual arm robots for live working in distribution networks,aiming to develop a highly autonomous and user-friendly live working robot for distribution networks.The main research content and innovative achievements are as follows:(1)In response to the low efficiency of distribution network robots in pure remote operation,low success rate and safety in autonomous operation,and high dependence on fixed distribution network scenarios,a human-machine collaborative high-level system architecture for distribution network live working robots is proposed,and the design and optimization of robot software and hardware are completed.A variety of intelligent work tools and mechanical quick change devices have been designed,which,under dual arm collaborative control,enrich the operation functions of the distribution network live working robot and improve its application value.By designing a trajectory library for robot tool replacement,limited trajectory units are arranged and combined,achieving rapid tool retrieval and transfer between arms,and improving robot operation efficiency.(2)A multimodal feedback teleoperation control strategy is proposed to address the issue of insufficient perception of the surrounding environment and the robot itself during teleoperation in outdoor scenarios.The use of VR helmets instead of flat screens has solved the problem of low visibility of screen video images under strong outdoor light.Design the pan tilt camera on the remote robot to rotate with the VR helmet,and add a digital twin 3D virtual robot from the helmet perspective,enhancing the operator’s perception of the surrounding environment and the robot’s own state.The visual feedback of virtual and real fusion combined with force sensing feedback based on end force sensors forms multimodal teleoperation.A bilateral hybrid teleoperation control strategy is proposed to address the issue of expanding the workspace of the master hand without changing the teleoperation resolution during master-slave heterogeneous teleoperation control.Improved the limitations of traditional bilateral remote operation control that require matching the workspace of heterogeneous master-slave ends,and can effectively complete remote operation of shaft hole assembly operations.(3)In response to the problem of limited public datasets and single detection targets in distribution network scenarios,a virtual synthesized dataset is proposed for pre training,and then the weights are transferred to the small sample real dataset to fine-tune the model,achieving the goal of accelerating training speed and improving detection accuracy.In response to the problem of weak perception ability of distribution network robots to the surrounding environment,the YOLOv7 detection network and Mask RCNN segmentation network were used to achieve rapid detection of multiple targets on distribution towers and accurate segmentation of power lines,respectively.Based on the detection and segmentation results,the triangulation principle of a binocular stereo camera is used to calculate the relative distance between the robot and the target,improving the robot’s autonomous perception ability.(4)Verification of the live working robot for human-machine collaboration in the actual distribution network for hot work.In practical work,in response to the problem of inaccurate localization and low efficiency when operators adjust the working position of the insulated bucket of the bucket arm truck through naked eye observation,an insulation bucket auxiliary guidance strategy based on deep learning power-line segmentation results and binocular stereo camera localization is proposed.By using machine vision to detect the distance between the robot and the power-line,the efficiency of insulation bucket adjustment is improved.During the live working process,in response to the complex challenges of live working steps,the division of labor between the two arms was coordinated.Based on the results of deep learning power line segmentation and the location of the stereo camera,power line extraction and attitude estimation were achieved.In the simulation,the coordination planning of the two arm robot was completed and sent to the physical robot for execution.Improve overall operational efficiency through visual localization,robotic arm planning,and tool execution linkage.