Research On Collision Avoidance Method Of Mechanical Arm Based On Visual Perception

In recent years,robot arm man-machine cooperation has been widely used in intelligent manufacturing,logistics warehousing,construction handling and other fields,in which the robot arm collision avoidance technology can effectively avoid the collision between the robot arm and the surrounding environment or obstacles,ensure the safety of equipment and operators,and improve the range of use and application scenarios of the robot arm.In view of the lack of obstacle perception of human body in the working space and off-line planning of robot arm’s trajectory task in traditional man-machine collaborative collision avoidance technology,this paper uses multiple depth cameras to construct a three-dimensional point cloud model of human body in the working space and establish a collision detection model between human and robot arm.An improved path planning algorithm is proposed to realize the path planning of the manipulator in the workspace so as to complete the task of collision avoidance.Firstly,in view of the problems such as long reconstruction time,high hardware cost and complex operation of traditional human 3D model construction methods,adaptive segmentation method was adopted,through filtering and statistical filtering of human body point cloud extraction and preprocessing was completed by using human body junction junction syncytin grid method provided in the software kit.Aiming at the problem of low similarity of human body point cloud information extracted from three perspectives in the actual industrial control environment,a two-step method of coarse registration and fine registration was adopted to complete the registration work.The experimental results show that this method can effectively realize the human point cloud data splicing from three perspectives to complete the construction of human point cloud model.Secondly,when enveloping human point cloud model is processed by axial enveloping box algorithm,there will be obvious gap between the enveloping box and human body,resulting in low accuracy of collision detection between robotic arm and human body.The minimum bounding box algorithm is used to construct the Gaussian sphere based on the convex hull of the object.According to the geometric constraints,the intersection point of the plane normal vector of the bounding box and the arc of the Gaussian sphere is solved to complete the updating of the bounding box volume.In order to improve the time complexity of the minimum bounding box algorithm,the genetic algorithm is introduced,using gene coding,selection,crossover,mutation and other operations.In order to complete the construction of minimum bounding box,the individuals in the population were inherited and optimized.The experimental results show that the improved minimum bounding box algorithm can keep the envelope tight and reduce the construction time of human point cloud model.Then,aiming at the problems such as low planning efficiency,long exploration time and tortuous planning path of traditional RRT* and bidirectional RRT* planning algorithms in complex environments,a path planning algorithm of IBRRT* manipulator based on redundant node filtering mechanism was proposed.Based on the IBRRT*programming algorithm,a local node replacement mechanism was introduced to avoid redundant expansion of nodes.The initial path from the starting point to the target point was explored for the first time,and the sampling area was constrained to eliminate the edge nodes one by one.Experimental results show that the proposed algorithm has fast convergence speed and high efficiency,and is feasible in practical environment.Finally,ROS actually built a robot arm collision avoidance system based on visual perception,realized human point cloud data acquisition,human point cloud model collision avoidance encircling box construction,robot arm trajectory planning and other functions.The experimental results verify the effectiveness of the relevant methods in this paper,which can effectively ensure the safety of human body in the process of humanmachine cooperation collision avoidance.