Optimistic Multi-Agent Policy Gradient

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhao, Wenshuai, Zhao, Yi, Li, Zhiyuan, Kannala, Juho, Pajarinen, Joni
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915551337512960
author Zhao, Wenshuai
Zhao, Yi
Li, Zhiyuan
Kannala, Juho
Pajarinen, Joni
author_facet Zhao, Wenshuai
Zhao, Yi
Li, Zhiyuan
Kannala, Juho
Pajarinen, Joni
contents *Relative overgeneralization* (RO) occurs in cooperative multi-agent learning tasks when agents converge towards a suboptimal joint policy due to overfitting to suboptimal behavior of other agents. No methods have been proposed for addressing RO in multi-agent policy gradient (MAPG) methods although these methods produce state-of-the-art results. To address this gap, we propose a general, yet simple, framework to enable optimistic updates in MAPG methods that alleviate the RO problem. Our approach involves clipping the advantage to eliminate negative values, thereby facilitating optimistic updates in MAPG. The optimism prevents individual agents from quickly converging to a local optimum. Additionally, we provide a formal analysis to show that the proposed method retains optimality at a fixed point. In extensive evaluations on a diverse set of tasks including the *Multi-agent MuJoCo* and *Overcooked* benchmarks, our method outperforms strong baselines on 13 out of 19 tested tasks and matches the performance on the rest.
format Preprint
id arxiv_https___arxiv_org_abs_2311_01953
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Optimistic Multi-Agent Policy Gradient
Zhao, Wenshuai
Zhao, Yi
Li, Zhiyuan
Kannala, Juho
Pajarinen, Joni
Machine Learning
Multiagent Systems
*Relative overgeneralization* (RO) occurs in cooperative multi-agent learning tasks when agents converge towards a suboptimal joint policy due to overfitting to suboptimal behavior of other agents. No methods have been proposed for addressing RO in multi-agent policy gradient (MAPG) methods although these methods produce state-of-the-art results. To address this gap, we propose a general, yet simple, framework to enable optimistic updates in MAPG methods that alleviate the RO problem. Our approach involves clipping the advantage to eliminate negative values, thereby facilitating optimistic updates in MAPG. The optimism prevents individual agents from quickly converging to a local optimum. Additionally, we provide a formal analysis to show that the proposed method retains optimality at a fixed point. In extensive evaluations on a diverse set of tasks including the *Multi-agent MuJoCo* and *Overcooked* benchmarks, our method outperforms strong baselines on 13 out of 19 tested tasks and matches the performance on the rest.
title Optimistic Multi-Agent Policy Gradient
topic Machine Learning
Multiagent Systems
url https://arxiv.org/abs/2311.01953