Vision-Based Obstacle Separation for Strawberry Harvesting in Clusters Using Hierarchical Reinforcement Learning

2026-07-15

AI Summary

In dense strawberry clusters, mature fruits are often occluded by unripe ones, leading to low success rates in direct harvesting. To address this challenge, this work proposes VGPA, a hierarchical reinforcement learning framework that decouples the task into two stages: obstacle separation and target grasping. The approach innovatively integrates a vision-guided high-level option selection mechanism with a low-level progressive adaptive exploration strategy (PAES), significantly improving policy convergence speed, exploration efficiency, and training stability. The method enables effective sim-to-real transfer, achieving a 96.7% success rate in simulation and 71.7%–88.3% in real-world experiments—substantially outperforming baseline methods—with only a marginal increase of approximately 1.22 seconds in average execution time.

Abstract

Selective harvesting in clustered strawberry environments is challenging because ripe fruits are often occluded by surrounding unripe fruits, making direct grasping unreliable. To address this problem, this paper proposes a hierarchical reinforcement learning framework, termed VGPA, which integrates a vision-guided decision mechanism and a Progressive Adaptive Exploration Strategy (PAES) for vision-based obstacle separation and harvesting. The task was decomposed into two sequential stages: obstacle separation and target grasping. At the high level, the vision-guided mechanism improved option selection and accelerated policy convergence. At the low level, PAES improved exploration efficiency and training stability during continuous control learning. In simulation experiments, the learned policy achieved a success rate of 96.7%. In addition, sim-to-real transfer experiments on a self-developed parallel robot showed that the proposed method achieved success rates ranging from 71.7% to 88.3%, outperforming direct picking while requiring only 1.22~s more average harvesting time. These results verified the effectiveness, generalization ability, and practical potential of the proposed method for robotic harvesting in complex clustered environments.

Problem

• strawberry harvesting
• obstacle separation
• clustered environment
• occlusion
• selective harvesting

Innovation

• Hierarchical Reinforcement Learning
• Vision-Guided Decision
• Obstacle Separation
• Progressive Adaptive Exploration Strategy
• Sim-to-Real Transfer