Towards Hierarchical Task Decomposition using Deep Reinforcement Learning for Pick and Place Subtasks

TLDR

Deep Reinforcement Learning offers adaptive robotic behaviors but requires massive data, and Learning from Demonstrations mitigates this yet demands many expert demos that are hard to obtain. The study proposes a multi‑subtask reinforcement learning framework that decomposes complex pick‑and‑place tasks into low‑level subtasks to reduce data requirements. The approach trains expert‑network subtasks with DRL, then assembles them via a high‑level choreographer, and evaluates the system in a pick‑and‑place simulator, outperforming an LfD benchmark in sample efficiency. The method outperforms an LfD benchmark in sample efficiency in simulation and transfers to a real robot, achieving robust grasping of various geometrically shaped objects.

Abstract

Deep Reinforcement Learning (DRL) is emerging as a promising approach to generate adaptive behaviors for robotic platforms. However, a major drawback of using DRL is the data-hungry training regime that requires millions of trial and error attempts, which is impractical when running experiments on robotic systems. Learning from Demonstrations (LfD) has been introduced to solve this issue by cloning the behavior of expert demonstrations. However, LfD requires a large number of demonstrations that are difficult to be acquired since dedicated complex setups are required. To overcome these limitations, we propose a multi-subtask reinforcement learning methodology where complex pick and place tasks can be decomposed into low-level subtasks. These subtasks are parametrized as expert networks and learned via DRL methods. Trained subtasks are then combined by a high-level choreographer to accomplish the intended pick and place task considering different initial configurations. As a testbed, we use a pick and place robotic simulator to demonstrate our methodology and show that our method outperforms a benchmark methodology based on LfD in terms of sample-efficiency. We transfer the learned policy to the real robotic system and demonstrate robust grasping using various geometric-shaped objects.