Research On Intelligent Detection And Grasping Planning Of Deformable Linear Objects

Flexible bodies,especially deformable linear objects(DLOs),such as wires,cables,ropes,etc.,are commonly used in manufacturing,food industry and household services.Among them,in the welding field of the 3C manufacturing industry,part of the process of welding various DLO parts(mainly including four processes: stripping,separating the line,arranging the cables in order,and soldering)still requires manual operations.This is because these linear flexible bodies are different from planar flexible bodies and body-shaped flexible bodies.The linear flexible bodies will not only produce extremely uncertain deformations under the action of external forces,but also complex crossings will occur between the flexible bodies.Winding state.This has led to the work of separating the wires and pressing the wires into the welding slot in a specific order in the existing welding equipment,which still needs to be completed manually,which has the defects of slow efficiency,low fault tolerance and uneven quality.Secondly,for wire-like linear flexible bodies,they have a certain degree of elasticity.When the external force is removed,they will pop out of the target position under the action of the elastic force generated by their own deformation.This will lead to wrong welding and missing welding in the subsequent welding process,leading to welding failure.Therefore,research on this kind of automatic welding robot for linear flexible wires is of great significance to the development of 3C manufacturing.Based on the above background,this paper proposes a linear flexible body intelligent detection and grasping planning algorithm based on deep learning to identify the cross state of flexible cables and plan the priority of cable grasping,and proposes A flexible wire assembly planning method based on visual feedback is used to deal with unexpected situations in which wire escapes from the target position under its own elastic force.In the welding process of flexible wires,several pretreatment steps are required before welding,that is,wire stripping,wire separation,wire sorting,and wire crimping into the welding groove.Among them,after the stripped wires pass through the separation mechanism,the wires will have different cross-shielding states under the action of their own elastic force,which will interfere with the subsequent process of inserting each wire into the corresponding welding slot by the clamping jaw.Therefore,in the visual inspection part,not only the wires need to be identified and located,but also the grabbing sequence needs to be planned and sorted accordingly to avoid the interference of the grabbing caused by the crossing of wires.This paper first uses the Mask R-CNN detection algorithm based on deep learning to identify and segment each wire,and obtain the cross-interference between the wires by analyzing the number of connected domains generated by segmenting each wire.Subsequently,a visual inspection algorithm based on YOLOv3 was used for comparison,and the intersection of all wires was obtained by directly identifying the end points and intersections of the wires.By comparing and analyzing the advantages and disadvantages of the two detection algorithms,it is determined that YOLOv3 is finally used as the visual detection algorithm in this paper.Based on the YOLOv3 network,the grasping priority of each wire is obtained through an experience-based grasping sequence algorithm.Finally,in the acquisition of the position and pose of the grasping point on the wire,the position coordinates of the grasping point on the wire are obtained by the area growing method.In the part of wire assembly planning,this paper adopts the DLO assembly planning method based on visual feedback.For the unexpected situations that may occur during the cable grabbing process(the cable slips from the fixture when it is grabbed by the fixture and moves to the welding slot,and the cable is grabbed into the welding slot due to its own elastic force.The phenomenon of popping in the welding groove),the DLO assembly planning method based on the area arrival control theory and visual feedback is designed to adjust the grabbing sequence of the cables,and the entire grabbing is realized in the face of accidents during the grabbing process The automation of the process and feedback process also effectively avoids the impact of detection jitter in visual inspection.Finally,this paper builds the hardware platform required for the experiment and designs the experimental process.Through the grasping experiment of DLO in different levels of complexity,the grasping planning algorithm based on the YOLOv3 network and the effectiveness of the DLO assembly planning method are verified.