Research Of Redundant Robot Dynamic Obstacle Avoidance Path Planning With Remote Operation In Augmented Reality

With the increasing use of robots, low degree of freedom robot and mobile robot can not complete some complicated tasks, in that time need redundancy robot to utilize self-motion in order to implement the first object (terminal pose) at the same time to meet a variety of secondary goal. The dynamic obstacle avoidance method is becoming more and more important part of the autonomous path planning. This paper watch the complex environment around robot with binocular vision and consider real-time of dynamic obstacle avoidance, recognizing and tracking dynamic obstacles, the redundant robot dynamic obstacle avoidance path planning method researched. The main content of this thesis includes the following aspects:First of all, the moving object recognition and tracking method are researched. The background for the rest of the scene, the frame differential method for dynamic target recognition, the use of adjacent two frame image pixel difference, connectivity analysis, judge connected area is greater than a given threshold, complete moving target recognition and calculate the target area. Use characteristic of the fusion of Cam shift mean shift algorithm for target tracking and extract the target area color characteristics and gradient characteristics, and calculate the probability density function of the candidate target and pap distance value, complete moving target tracking.Secondly, the three-dimensional location method of dynamic object is researched. Analysis of binocular vision model and parallax principle, the use of target edge feature point matching calculation parallax information, complete the targets of three dimensional positioning. According to the moving target tracking technique, a kind of double channel parallel binocular vision target identification and positioning method, a channel in the target area for extraction and tracking, another channel edge feature point matching and calculating the average target3d coordinates. Dual channel parallel processing reached the updated moving target3d information purpose.Thirdly, a pre-selected minimum distance index is presented considering the real-time performance in dynamic obstacle avoidance for redundant robot. Using OBB (Oriented Bounding Box) bounding box impact robot model, getting rid of safety link by intersection test, the real-time minimum distance is calculated by local mapping distance calculation method in link coordinate system. Based on that, avoidance factor and escaping velocity is built related to minimum distance and then zero space mapping matrix is utilized to accomplish dynamic path planning. Compared with traditional path planning method, the optimization method reduced15.5%in the overall planning time.Finally, Built based on binocular vision redundant robot dynamic obstacle avoidance path planning experiment platform, verify the target recognition and tracking method, the three dimensional positioning method is correct, preliminary selection minimum distance index of the redundant robot dynamic obstacle avoidance method is effective and high efficiency, and improve the safety performance of the robot.