Robot Learning of Shifting Objects for Grasping in Cluttered Environments

TLDR

Robotic grasping in cluttered environments is often infeasible because obstacles block grasps, making pre‑grasping manipulation such as shifting or pushing necessary. The study presents an algorithm that learns to shift objects to increase grasp probability, including optimal manipulation‑primitive poses and non‑prehensible actions that explicitly raise grasp success. The algorithm is trained self‑supervised on approximately 25 000 grasp and 2 500 shift actions and applied to bin‑picking, enabling complete bin emptying. The approach yields data‑efficient learning, achieves complete bin emptying, records 274 ± 3 picks per hour, and generalizes to novel objects.

Abstract

Robotic grasping in cluttered environments is often infeasible due to obstacles preventing possible grasps. Then, pre-grasping manipulation like shifting or pushing an object becomes necessary. We developed an algorithm that can learn, in addition to grasping, to shift objects in such a way that their grasp probability increases. Our research contribution is threefold: First, we present an algorithm for learning the optimal pose of manipulation primitives like clamping or shifting. Second, we learn non-prehensible actions that explicitly increase the grasping probability. Making one skill (shifting) directly dependent on another (grasping) removes the need of sparse rewards, leading to more data-efficient learning. Third, we apply a real-world solution to the industrial task of bin picking, resulting in the ability to empty bins completely. The system is trained in a self-supervised manner with around 25 000 grasp and 2500 shift actions. Our robot is able to grasp and file objects with 274±3 picks per hour. Furthermore, we demonstrate the system's ability to generalize to novel objects.