A Visual Reinforcement Learning-Based Separate Primitive Policy for Peg-in-Hole Tasks
This work addresses the challenge of efficient assembly in robotics, offering a method that enhances learning performance for peg-in-hole tasks.
The paper proposes a Separate Primitive Policy (S2P) for visual reinforcement learning in peg-in-hole tasks, enabling agents to learn location and insertion actions simultaneously. S2P improves sample efficiency and success rate under force constraints, as demonstrated in simulation and real-world experiments.
For peg-in-hole tasks, humans rely on binocular visual perception to locate the peg above the hole surface and then proceed with insertion. This paper draws insights from this behavior to enable agents to learn efficient assembly strategies through visual reinforcement learning. Hence, we propose a Separate Primitive Policy (S2P) to learn how to derive location and insertion actions simultaneously. S2P is compatible with model-free reinforcement learning algorithms. Ten insertion tasks featuring different polygons are developed as benchmarks for evaluations. Simulation experiments show that S2P can boost the sample efficiency and success rate even with force constraints. Real-world experiments are also performed to verify the feasibility of S2P. Ablations are finally given to discuss the generalizability of S2P and some factors that affect its performance.