A Study on Long-Close Distance Coordination Control Strategy for Litchi Picking

For the automated robotic picking of bunch-type fruit, strategy is to roughly determine location bunches, plan route from a remote location, and then locate point precisely at more appropriate, closer location. The latter can reduce amount information be processed obtain precise detailed features, thus improving accuracy vision system. In this study, long-close distance coordination control for litchi robot was proposed based on an Intel Realsense D435i camera combined with cloud map collected by camera. YOLOv5 object detection network DBSCAN clustering method were used bunch fruits long deduce sequence picking. After reaching close-distance position, Mask RCNN instance segmentation segment distinctive bifurcate stems in field view. By processing masks, dual reference model “Point + Line” proposed, which guided arm. Compared existing studies, took into account advantages disadvantages depth cameras. experimenting complete process, density-clustering approach able classify different bunches distance, while success rate 88.46% achieved during fruit-bearing branch locating. This exploratory work that provided theoretical technical future research fruit-picking robots.