Obstacle Avoidance Of Manipulator In Unknown Environment Based On Binocular Vision

With the continuous development of artificial intelligence and automation technology,more and more applications need to strengthen the intelligent manipulator.Based on RBT-6T/S03 S Six-DOF robots,and the obstacle detection and obstacle avoidance path planning of the manipulator in unknown environment are studied in this paper.The main contents of this paper include the establishment of Six-DOF robots' model,kinematics analysis,determination of obstacle position in the environment by binocular camera,hand-eye calibration between the manipulator and binocular camera,traditional RRT algorithm and improved RRT algorithm for obstacle avoidance path planning.The details are as follows:Firstly,the structure of RBT-6T/S03 S robot is introduced.According to the structure sketch of the manipulator,the D-H coordinate system is established.In order to solve the complex problem of inverse kinematics calculation,BP neural network is used to analyze the inverse kinematics solution of the manipulator.The input is the parameters of the end pose matrix,and the output is the angle values of each joint.The training and testing samples are obtained by recording the reposition motion of the manipulator.Compared with the actual value,the predicted result is close to the actual value.In order to optimize the solution,the genetic algorithm is used to optimize the BP neural network.The flow chart and implementation method of the algorithm are given.The results show that the prediction accuracy is obviously improved.Secondly,the principle of camera imaging is introduced.The camera is calibrated by the camera calibration toolbox of MATLAB.Then the three-dimensional coordinates of obstacle feature points are solved by Structure from Motion algorithm,and the position of obstacle is determined.The main steps of SFM algorithm are as follows: 1.Extracting feature points and matching.2.Calculating the basic matrix,and then getting the essential matrix.3.Calculating the motion parameters of the camera and obtaining the rotation and translation matrix.4.According to the internal parameters and rotation translation matrix,the projection matrix is obtained,and the three-dimensional coordinates of the points are determined.Then the hand-eye calibration that "Eye-to-hand" is introduced,and the calibration process and results are given.Finally,in this paper,obstacle avoidance algorithm of manipulator is studied,collision detection method between manipulator and obstacle is analyzed,and the rapidly-exploring random tree(RRT)algorithm is used to plan obstacle avoidance path.Aiming at the complex computation problem of Cartesian space,this paper chooses collision-free path planning in joint space.In the traditional RRT algorithm,the planned path is too random.This paper introduces the method of target bias and dynamically changes the step size of path search to realize RRT algorithm.The simulation results show that the improved RRT is superior to the traditional algorithm in planning time and path.Then,the feasibility of the improved RRT algorithm is verified by experiments on the manipulator body.