Six‐DoF Strawberry Harvesting Pose Selection and Adaptive Obstacle Clearing in Occluded Environments

Abstract

In complex, unstructured orchard environments, accurate end‐effector pose selection is essential for low‐damage strawberry harvesting. This study proposes an occlusion‐aware 6‐DoF harvesting pose‐selection strategy with adaptive obstacle clearing for a fixed‐stroke two‐finger end‐effector. Building on an existing 5‐DoF fruit pose, the method generated a finite set of candidate γ‐ angles around the estimated fruit longitudinal axis. Each candidate was evaluated using a ripeness‐weighted obstruction‐risk score, and the feasible candidate with the minimum score was selected. RRT‐Connect was employed for reachability and collision‐feasibility verification of the selected pose. When a collision‐free approach cannot be obtained, adaptive obstacle clearing was activated according to local obstacle distribution and fruit ripeness. Field experiments showed that, compared with the 5‐DoF baseline, the complete strategy increased harvesting success rate from 65.5% to 78.0% and reduced damage rate from 7.5% to 5.5%, while increasing execution time from 5.55 to 6.57 s. These results demonstrated that the proposed 6‐DoF pose‐selection strategy combined with adaptive obstacle clearing improved harvesting reliability under occluded conditions.