Research On Robot Collision Avoidance Method Based On Visual Perception

With the widespread application of robots in the manufacturing field,many scenarios that require collaboration between humans and robots have emerged.The participation of robots in human tasks can not only ensure work efficiency but also improve product quality.In the context of human-robot collaboration,how to ensure human safety is a crucial issue.This article focuses on the topic of robot collision avoidance technology in a human-machine collaborative environment,and conducts research on robot collision avoidance methods based on visual perception,taking human body data detection and robotic arm collision avoidance path planning in the process of human-machine collaboration as the research objects,a visual perception robot collision avoidance system based on the ROS(Robotic Operation System)robot development platform is constructed.The main tasks completed in this paper are as follows:First of all,an adaptive image segmentation algorithm is proposed for the human body data of the 3D point cloud collected by the Kinect2.0 camera in the process of human-machine collaboration.According to the coordinates of the human body center in the three-dimensional space calculated according to the Kinect2.0 bone data,the threshold of the straight-through filter is adaptively modified,and the statistical filtering algorithm is combined to realize the threshold segmentation and discrete point processing of the three-dimensional point cloud data of the human body,and is completed by the ICP(Iterative Closest Point)algorithm Mosaic processing of human body 3D point cloud data.Experiments show that this method can identify the three-dimensional point cloud human body model markers in complex scenes,and can accurately obtain complete three-dimensional point cloud human body data.Secondly,aiming at the collision detection problem with the human body during the motion of the robotic arm,a hierarchical bounding box algorithm based on the AABB(Axis-aligned bounding box)bounding box is designed,and the acquired complete human body 3D point cloud data is enveloped.Through the research of collision detection models such as bounding boxes in the spatial domain,the advantages and disadvantages of each bounding box are analyzed,and finally,the AABB bounding box is used to design a hierarchical bounding box algorithm that can envelop human body data.Then,aiming at the trajectory planning problem of the manipulator,a collision avoidance trajectory planning algorithm with an improved bidirectional and fast expanding random tree is proposed.By introducing the target bias strategy in the RRT-Connect algorithm,the sampling point is guided to expand toward the target point,and the step length is changed during the expansion of the random tree to speed up the expansion of the random tree.Finally,the pruning process of the random tree is completed by the greedy algorithm.Simulation experiments show that the improved algorithm can not only reduce the path generation time,but also shorten the path length relatively.Finally,a visual perception robot collision avoidance system based on the ROS robot development platform is constructed,which realizes the functions of image acquisition,data processing and analysis,and robot collision avoidance trajectory planning.The visual perception robot collision avoidance method designed by the ROS development platform is verified,and the experimental results verify the effectiveness of the algorithm proposed in this paper.Therefore,the robot collision avoidance method based on visual perception studied in this paper has achieved the expected effect and can be applied in the process of human-machine collaboration to ensure the safety of the human body.