Sample-Efficient Reinforcement Learning with Temporal Logic Objectives: Leveraging the Task Specification to Guide Exploration

TL;DR

This paper introduces a novel reinforcement learning algorithm that leverages temporal logic task specifications to guide exploration, significantly improving sample-efficiency especially in complex or large MDPs.

Contribution

The paper presents a new task-driven exploration strategy for RL with temporal logic objectives, enhancing learning speed and efficiency over existing methods.

Findings

The proposed method achieves faster learning than competitive approaches.

Sample-efficiency improves as task complexity or MDP size increases.

Theoretical analysis supports the effectiveness of the exploration strategy.

Abstract

This paper addresses the problem of learning optimal control policies for systems with uncertain dynamics and high-level control objectives specified as Linear Temporal Logic (LTL) formulas. Uncertainty is considered in the workspace structure and the outcomes of control decisions giving rise to an unknown Markov Decision Process (MDP). Existing reinforcement learning (RL) algorithms for LTL tasks typically rely on exploring a product MDP state-space uniformly (using e.g., an $\epsilon$-greedy policy) compromising sample-efficiency. This issue becomes more pronounced as the rewards get sparser and the MDP size or the task complexity increase. In this paper, we propose an accelerated RL algorithm that can learn control policies significantly faster than competitive approaches. Its sample-efficiency relies on a novel task-driven exploration strategy that biases exploration towards directions that may contribute to task satisfaction. We provide theoretical analysis and extensive comparative experiments demonstrating the sample-efficiency of the proposed method. The benefit of our method becomes more evident as the task complexity or the MDP size increases.