Research On Path Planning Method Of Apple Picking Robot Based On Vision Detection

China is the largest country in apple production in the world.With the rising cost of apple picking,the use of picking robots instead of manual work can not only save costs,but also shorten the apple picking cycle,which has broad application prospects and important economic value.However,in a complex natural environment,the existing technology cannot make the working efficiency of the picking robot meet the actual production demand,and cannot reach the practical level.With the support of the Jiangsu Modern Agriculture Project(BE2020406),this thesis studies the key technologies of the picking robot from three aspects: the overall design of the robot system,the detection method of apple,and the path planning method of the manipulator.The main work is summarized as follows:Firstly,a scheme of the picking robot system is designed for the complex orchard working environment.The whole system includes the mobile chassis module,manipulator module,end grab module,vision module,and control module.According to the needs of apple picking,the hardware design and selection of each module are carried out,and the corresponding software development is completed.Secondly,the mathematical modeling and analysis of the picking robot system are carried out.The kinematics model of the manipulator is established and the simulation experiment is carried out;The camera model and visual calibration method are studied,the conversion relationship between the camera coordinate system and the manipulator coordinate system is obtained,and finally the rotation angle of the manipulator’s joint is obtained to realize visual servo control.Thirdly,an apple detection method based on the Des-YOLO v4 algorithm is proposed.Firstly,based on the YOLO v4 network,a network structure named Des-YOLO is designed to improve the detection speed of the algorithm;Then soft NMS is used instead of NMS for non-maximum suppression to improve the detection accuracy;Finally,a category loss function based on AP-Loss is proposed to improve the accuracy of the proposed algorithm.The experimental results show that the improved algorithm has better performance,with a m AP of 93.1% and a detection speed of 53 f/s.The picking robot can realize the rapid and accurate detection of apples in complex environment.Fourthly,the collision-free path planning method of picking manipulators is studied.By analyzing the basic principles and characteristics of the RRT(Rapidly-exploring Random Tree)algorithm and RRT-connect algorithm,an improved planning algorithm called hybrid RRT is designed.This algorithm obtains the initial path through dual-tree search,then combines the initial trees and target trees of dual-tree search into one,and conducts informed sampling of the combined single tree to find a near-optimal path.The simulation results show that the proposed algorithm improves the efficiency and quality of picking robot path planning.Finally,the apple picking robot experimental platform is built in the simulated orchard environment in the laboratory,and the apple picking experiment of the robot system is completed.The experimental results verify the effectiveness of the proposed method.Compared with the previous picking robots,the picking success rate of the picking robot system designed in this thesis has reached 93.7%,and its working effect has basically reached the ideal level,which promotes the practicability and industrialization of domestic apple picking robots.