Design And Research Of Obstacleavoidance System For Apple Picking Manipulator

Fruit harvesting is an important part of agricultural production.The development and application of agricultural intelligent picking robots can solve the current shortage of labor in China,greatly reduce production costs,and improve market competitiveness.Although the development of industrial robots is becoming more and more mature now,the agricultural picking environment is more complicated and the requirements for path search are higher.In this paper,the obstacle avoidance system is researched with apples as the object of picking.First,according to the growth characteristics of the apple tree,the structure of the picking robot is divided into three parts: positioning mechanism,posture adjustment mechanism and a picking mechanism.A five-degree-of-freedom apple picking mechanism is obtained through a combination of configuration and size.Then,the mathematical model of the robot is established based on the D-H parameter method,and the forward solution and the calculation of the working space are performed.At the same time,in order to simplify the calculation of the inverse solution,the inclination angle of the axis of the apple is solved,and the joint angle is preset.In terms of path search,researching and analysing on common "random" algorithms find that the "node" generated by traditional algorithms in path search has large randomness and blindness.For this reason,an improved RRT-connect algorithm is proposed,which can improves the efficiency and stability of path search to a certain extent.Then,an octree algorithm is used to establish an obstacle model for the point cloud data of the picking environment obtained by the vision system.Finally,the improved RRT-connect algorithm is used to complete the path search task under the environment model built by the octree.In terms of trajectory planning,the trajectory planning algorithms in Cartesian coordinate system and joint coordinate system are studied firstly,and it is found that the fifth-order polynomial trajectory planning in joint coordinate system has a small calculation amount and relatively stable motion.Then,considering the results of the path search,it is necessary to plan the speed of the intermediate nodes,and propose a step-by-step trajectory planning idea.Research and analyze common speed planning methods,and use them in conjunction with quintic polynomial planning methods to generate speed and acceleration curves.By comparing several sets of curves,it is found that the quintic polynomial trajectory planning algorithm based on S-shaped curves has good stability.The collision detection method is studied.The ball and box bounding boxes are used to simplify apples and branches,respectively,and the collision between the robot arm link and the apple and branch trunks was converted into the collision problem between the cylinder and the ball and cylinder.The conditions of the collision.Then,the genetic algorithm is applied to the multi-fruit picking problem,thereby obtaining an optimal set of picking orders.Finally,calculate the axis of Apple's point cloud data to determine the end posture of the picking equipment.