POAR: Efficient Policy Optimization via Online Abstract State Representation Learning

TL;DR

POAR introduces an integrated approach combining RL and SRL with dynamic loss weighting and expert demonstrations, significantly improving sample efficiency and performance in high-dimensional robotic tasks.

Contribution

The paper presents a novel algorithm that seamlessly integrates SRL into policy optimization, addressing overfitting and enhancing learning efficiency in complex environments.

Findings

POAR outperforms state-of-the-art RL algorithms in sample efficiency and rewards.

It effectively handles high-dimensional tasks and enables training real robots from scratch.

The method leverages expert demonstrations and real-time state graph monitoring.

Abstract

While the rapid progress of deep learning fuels end-to-end reinforcement learning (RL), direct application, especially in high-dimensional space like robotic scenarios still suffers from low sample efficiency. Therefore State Representation Learning (SRL) is proposed to specifically learn to encode task-relevant features from complex sensory data into low-dimensional states. However, the pervasive implementation of SRL is usually conducted by a decoupling strategy in which the observation-state mapping is learned separately, which is prone to over-fit. To handle such problem, we summarize the state-of-the-art (SOTA) SRL sub-tasks in previous works and present a new algorithm called Policy Optimization via Abstract Representation which integrates SRL into the policy optimization phase. Firstly, We engage RL loss to assist in updating SRL model so that the states can evolve to meet the demand of RL and maintain a good physical interpretation. Secondly, we introduce a dynamic loss weighting mechanism so that both models can efficiently adapt to each other. Thirdly, we introduce a new SRL prior called domain resemblance to leverage expert demonstration to improve SRL interpretations. Finally, we provide a real-time access of state graph to monitor the course of learning. Experiments indicate that POAR significantly outperforms SOTA RL algorithms and decoupling SRL strategies in terms of sample efficiency and final rewards. We empirically verify POAR to efficiently handle tasks in high dimensions and facilitate training real-life robots directly from scratch.