Picking Trajectory Planning of Citrus Based on Improved Immune Algorithm and Binocular Vision

Abstract

Aiming at the phenomenon that to pick citrus by robot arm is in low efficiency, this paper chooses improved immune algorithm (IIA) for picking path planning of the citrus fruits on the surface of the tree canopy after many experiments. IIA is proposed by improving the neighborhood structure of basic immune algorithm (BIA) and using tabu search strategy to search the neighborhood structure of the current optimal solution which is already got by immune search in the final stage intensively. The world coordinates of citrus fruits are obtained by processing the photos taken by a ZED camera based on the principle of binocular vision. The experiment results show that when picking 6, 20 and 31 citrus fruits, the average planning time of IIA are 13.33%, 21.49% and 23.96% less than BIA, and the average picking distance are 0%, 0.66% and 0.67% shorter than BIA. This shows that IIA can not only effectively shorten the time of trajectory planning, but also shorten the distance of picking path, which provides theoretical support of improving the working efficiency of picking robot.