Probabilistically Guaranteed Satisfaction of Temporal Logic Constraints During Reinforcement Learning

TL;DR

This paper introduces a new reinforcement learning method that guarantees probabilistic satisfaction of temporal logic constraints during learning, ensuring constraints are met in each episode rather than after many episodes.

Contribution

It presents an automata-theoretic approach to ensure probabilistic constraint satisfaction in each episode, differing from traditional penalty-based methods, with theoretical guarantees and practical demonstration.

Findings

Guarantees probabilistic satisfaction of constraints in each episode.

Provides a lower bound on the probability of constraint satisfaction.

Demonstrates effectiveness in a drone navigation scenario.

Abstract

We propose a novel constrained reinforcement learning method for finding optimal policies in Markov Decision Processes while satisfying temporal logic constraints with a desired probability throughout the learning process. An automata-theoretic approach is proposed to ensure the probabilistic satisfaction of the constraint in each episode, which is different from penalizing violations to achieve constraint satisfaction after a sufficiently large number of episodes. The proposed approach is based on computing a lower bound on the probability of constraint satisfaction and adjusting the exploration behavior as needed. We present theoretical results on the probabilistic constraint satisfaction achieved by the proposed approach. We also numerically demonstrate the proposed idea in a drone scenario, where the constraint is to perform periodically arriving pick-up and delivery tasks and the objective is to fly over high-reward zones to simultaneously perform aerial monitoring.