Abstract Demonstrations and Adaptive Exploration for Efficient and Stable Multi-step Sparse Reward Reinforcement Learning

TL;DR

This paper introduces A^2, a novel DRL exploration method combining abstract demonstrations and adaptive exploration, significantly enhancing learning efficiency and stability in long-horizon, sparse reward tasks like robotic manipulation.

Contribution

The paper proposes A^2, a new exploration technique that decomposes tasks into subtasks and adaptively adjusts exploration, improving DRL performance on complex sparse reward tasks.

Findings

A^2 improves learning efficiency in robotic tasks.

A^2 enhances stability of DRL algorithms.

A^2 outperforms baseline methods in experiments.

Abstract

Although Deep Reinforcement Learning (DRL) has been popular in many disciplines including robotics, state-of-the-art DRL algorithms still struggle to learn long-horizon, multi-step and sparse reward tasks, such as stacking several blocks given only a task-completion reward signal. To improve learning efficiency for such tasks, this paper proposes a DRL exploration technique, termed A^2, which integrates two components inspired by human experiences: Abstract demonstrations and Adaptive exploration. A^2 starts by decomposing a complex task into subtasks, and then provides the correct orders of subtasks to learn. During training, the agent explores the environment adaptively, acting more deterministically for well-mastered subtasks and more stochastically for ill-learnt subtasks. Ablation and comparative experiments are conducted on several grid-world tasks and three robotic manipulation tasks. We demonstrate that A^2 can aid popular DRL algorithms (DQN, DDPG, and SAC) to learn more efficiently and stably in these environments.