Average-Reward Learning and Planning with Options

TL;DR

This paper extends the options framework for reinforcement learning to average-reward MDPs, introducing new algorithms and proving their convergence, with experiments demonstrating their effectiveness in a continuous domain.

Contribution

It introduces convergent off-policy inter-option learning algorithms and models for average-reward MDPs, extending the options framework and option-interrupting behavior.

Findings

Algorithms converge in average-reward settings

Effective in continuing Four-Room domain

Extends options framework to new setting

Abstract

We extend the options framework for temporal abstraction in reinforcement learning from discounted Markov decision processes (MDPs) to average-reward MDPs. Our contributions include general convergent off-policy inter-option learning algorithms, intra-option algorithms for learning values and models, as well as sample-based planning variants of our learning algorithms. Our algorithms and convergence proofs extend those recently developed by Wan, Naik, and Sutton. We also extend the notion of option-interrupting behavior from the discounted to the average-reward formulation. We show the efficacy of the proposed algorithms with experiments on a continuing version of the Four-Room domain.