Randomized Entity-wise Factorization for Multi-Agent Reinforcement Learning

Conference Paper (2021)
Author(s)

Shariq Iqbal (University of Oxford)

Christian A. Schroeder de Witt (University of Oxford)

Bei Peng (University of Oxford)

Wendelin Böhmer (TU Delft - Algorithmics)

Shimon Whiteson (University of Oxford)

Fei Sha

Research Group
Algorithmics
Copyright
© 2021 Shariq Iqbal, Christian A. Schroeder de Witt, Bei Peng, J.W. Böhmer, Shimon Whiteson, Fei Sha
More Info
expand_more
Publication Year
2021
Language
English
Copyright
© 2021 Shariq Iqbal, Christian A. Schroeder de Witt, Bei Peng, J.W. Böhmer, Shimon Whiteson, Fei Sha
Research Group
Algorithmics
Volume number
139
Pages (from-to)
4596-4606
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

Real world multi-agent tasks often involve varying types and quantities of agents and non-agent entities; however, agents within these tasks rarely need to consider all others at all times in order to act effectively. Factored value function approaches have historically leveraged such independences to improve learning efficiency, but these approaches typically rely on domain knowledge to select fixed subsets of state features to include in each factor. We propose to utilize value function factoring with random subsets of entities in each factor as an auxiliary objective in order to disentangle value predictions from irrelevant entities. This factoring approach is instantiated through a simple attention mechanism masking procedure. We hypothesize that such an approach helps agents learn more effectively in multi-agent settings by discovering common trajectories across episodes within sub-groups of agents/entities. Our approach, Randomized Entity-wise Factorization for Imagined Learning (REFIL), outperforms all strong baselines by a significant margin in challenging StarCraft micromanagement tasks.

Files

Iqbal21a_1_.pdf
(pdf | 2.9 Mb)
License info not available