Saved in:
Bibliographic Details
Main Authors: Marchesini, Enrico, Baisero, Andrea, Bhati, Rupali, Amato, Christopher
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.15381
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910596237099008
author Marchesini, Enrico
Baisero, Andrea
Bhati, Rupali
Amato, Christopher
author_facet Marchesini, Enrico
Baisero, Andrea
Bhati, Rupali
Amato, Christopher
contents Value factorization is a popular paradigm for designing scalable multi-agent reinforcement learning algorithms. However, current factorization methods make choices without full justification that may limit their performance. For example, the theory in prior work uses stateless (i.e., history) functions, while the practical implementations use state information -- making the motivating theory a mismatch for the implementation. Also, methods have built off of previous approaches, inheriting their architectures without exploring other, potentially better ones. To address these concerns, we formally analyze the theory of using the state instead of the history in current methods -- reconnecting theory and practice. We then introduce DuelMIX, a factorization algorithm that learns distinct per-agent utility estimators to improve performance and achieve full expressiveness. Experiments on StarCraft II micromanagement and Box Pushing tasks demonstrate the benefits of our intuitions.
format Preprint
id arxiv_https___arxiv_org_abs_2408_15381
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On Stateful Value Factorization in Multi-Agent Reinforcement Learning
Marchesini, Enrico
Baisero, Andrea
Bhati, Rupali
Amato, Christopher
Artificial Intelligence
Value factorization is a popular paradigm for designing scalable multi-agent reinforcement learning algorithms. However, current factorization methods make choices without full justification that may limit their performance. For example, the theory in prior work uses stateless (i.e., history) functions, while the practical implementations use state information -- making the motivating theory a mismatch for the implementation. Also, methods have built off of previous approaches, inheriting their architectures without exploring other, potentially better ones. To address these concerns, we formally analyze the theory of using the state instead of the history in current methods -- reconnecting theory and practice. We then introduce DuelMIX, a factorization algorithm that learns distinct per-agent utility estimators to improve performance and achieve full expressiveness. Experiments on StarCraft II micromanagement and Box Pushing tasks demonstrate the benefits of our intuitions.
title On Stateful Value Factorization in Multi-Agent Reinforcement Learning
topic Artificial Intelligence
url https://arxiv.org/abs/2408.15381