Harvest motion planning for mango picking robot based on improved RRT-Connect

Abstract

Aiming at the problems of long motion path planning time and low picking efficiency of picking robots in unstructured orchard environments, a heuristic dynamic Rapidly-exploring Random Tree Connect motion planning algorithm (HDRRT-Connect) for picking robots for fast mango harvesting path planning was proposed in this study. The algorithm was obtained by introducing adaptive target gravitation strategy and heuristic dynamic step strategy based on RRT-Connect algorithm. It adjusts the step-size according to the information of the orchard environment as well as the path searching situation, so as to avoid falling into the local optimum of the path. The prototype based on the algorithm was used to carry out picking experiments in the natural orchard environment. The prototype picking test under the natural environment of the orchard is carried out, and the test results showed that the average path cost of the HDRRT-Connect algorithm was 95.7739, the average planning time was 0.448 s, and the success rate was 90%. Compared with the RRT, RRT-Connect and Probabilistic Roadmaps (PRM) algorithms, the HDRRT-Connect planning speed was improved by 95%, 24% and 59%, respectively, and the path cost was reduced by 35%, 13% and 18%, respectively. The results of the experiment verified the feasibility and efficiency of the improved algorithm. The HDRRT-Connect algorithm proposed in this study could effectively shorten the planning time, reduce the search path cost and improve the planning success rate. The research provides technical support for the fast-harvesting operation of mango picking robot.