State Abstraction in MAXQ Hierarchical Reinforcement Learning

Abstract

Many researchers have explored methods for hierarchical reinforcement learning (RL) with temporal abstractions, in which abstract actions are defined that can perform many primitive actions before terminating. However, little is known about learning with state abstractions, in which aspects of the state space are ignored. In previous work, we developed the MAXQ method for hierarchical RL. In this paper, we define five conditions under which state abstraction can be combined with the MAXQ value function decomposition. We prove that the MAXQ-Q learning algorithm converges under these conditions and show experimentally that state abstraction is important for the successful application of MAXQ-Q learning.

7 pages, 2 figures

References in corpus (1)

• Hierarchical Solution of Markov Decision Processes using Macro-actions

Cited by in corpus (5)

• Hierarchical principles of embodied reinforcement learning: A review
• Learning Abstract Models for Strategic Exploration and Fast Reward Transfer
• Abstract Value Iteration for Hierarchical Reinforcement Learning
• Dynamic probabilistic logic models for effective abstractions in RL
• Goal Agnostic Planning using Maximum Likelihood Paths in Hypergraph World Models