Boosting Value Decomposition via Unit-Wise Attentive State Representation for Cooperative Multi-Agent Reinforcement Learning

In cooperative multi-agent reinforcement learning (MARL), the environmental stochasticity and uncertainties will increase exponentially when the number of agents increases, which puts hard pressure on how to come up with a compact latent representation from partial observation for boosting value dec...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2023-05
Hauptverfasser: Zhao, Qingpeng, Zhu, Yuanyang, Liu, Zichuan, Wang, Zhi, Chen, Chunlin
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Zhao, Qingpeng
Zhu, Yuanyang
Liu, Zichuan
Wang, Zhi
Chen, Chunlin
description In cooperative multi-agent reinforcement learning (MARL), the environmental stochasticity and uncertainties will increase exponentially when the number of agents increases, which puts hard pressure on how to come up with a compact latent representation from partial observation for boosting value decomposition. To tackle these issues, we propose a simple yet powerful method that alleviates partial observability and efficiently promotes coordination by introducing the UNit-wise attentive State Representation (UNSR). In UNSR, each agent learns a compact and disentangled unit-wise state representation outputted from transformer blocks, and produces its local action-value function. The proposed UNSR is used to boost the value decomposition with a multi-head attention mechanism for producing efficient credit assignment in the mixing network, providing an efficient reasoning path between the individual value function and joint value function. Experimental results demonstrate that our method achieves superior performance and data efficiency compared to solid baselines on the StarCraft II micromanagement challenge. Additional ablation experiments also help identify the key factors contributing to the performance of UNSR.
format Article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2813744388</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2813744388</sourcerecordid><originalsourceid>FETCH-proquest_journals_28137443883</originalsourceid><addsrcrecordid>eNqNzcsKwjAQBdAgCIr2HwKuCzWp2m194UI3PpcSZFqm1ExNpq78eKP4Aa6Ge--B6Yi-0nocZ6lSPRF5XyVJoqYzNZnovnjNiTyjLeXZ1C3IJdzo3pBHRrLyiUaeLHJ8QQ8yZwbL-AR5YMMg99A48KEyX1yQkwuiBpz5ol1bM8Z5GUCgaMN-g_snbcE4G34ORbcwtYfodwditF4dF5u4cfRowfO1otbZMF1VNtazNNVZpv9Tbzr7UB0</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2813744388</pqid></control><display><type>article</type><title>Boosting Value Decomposition via Unit-Wise Attentive State Representation for Cooperative Multi-Agent Reinforcement Learning</title><source>Free E- Journals</source><creator>Zhao, Qingpeng ; Zhu, Yuanyang ; Liu, Zichuan ; Wang, Zhi ; Chen, Chunlin</creator><creatorcontrib>Zhao, Qingpeng ; Zhu, Yuanyang ; Liu, Zichuan ; Wang, Zhi ; Chen, Chunlin</creatorcontrib><description>In cooperative multi-agent reinforcement learning (MARL), the environmental stochasticity and uncertainties will increase exponentially when the number of agents increases, which puts hard pressure on how to come up with a compact latent representation from partial observation for boosting value decomposition. To tackle these issues, we propose a simple yet powerful method that alleviates partial observability and efficiently promotes coordination by introducing the UNit-wise attentive State Representation (UNSR). In UNSR, each agent learns a compact and disentangled unit-wise state representation outputted from transformer blocks, and produces its local action-value function. The proposed UNSR is used to boost the value decomposition with a multi-head attention mechanism for producing efficient credit assignment in the mixing network, providing an efficient reasoning path between the individual value function and joint value function. Experimental results demonstrate that our method achieves superior performance and data efficiency compared to solid baselines on the StarCraft II micromanagement challenge. Additional ablation experiments also help identify the key factors contributing to the performance of UNSR.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Ablation ; Decomposition ; Multiagent systems ; Representations</subject><ispartof>arXiv.org, 2023-05</ispartof><rights>2023. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>776,780</link.rule.ids></links><search><creatorcontrib>Zhao, Qingpeng</creatorcontrib><creatorcontrib>Zhu, Yuanyang</creatorcontrib><creatorcontrib>Liu, Zichuan</creatorcontrib><creatorcontrib>Wang, Zhi</creatorcontrib><creatorcontrib>Chen, Chunlin</creatorcontrib><title>Boosting Value Decomposition via Unit-Wise Attentive State Representation for Cooperative Multi-Agent Reinforcement Learning</title><title>arXiv.org</title><description>In cooperative multi-agent reinforcement learning (MARL), the environmental stochasticity and uncertainties will increase exponentially when the number of agents increases, which puts hard pressure on how to come up with a compact latent representation from partial observation for boosting value decomposition. To tackle these issues, we propose a simple yet powerful method that alleviates partial observability and efficiently promotes coordination by introducing the UNit-wise attentive State Representation (UNSR). In UNSR, each agent learns a compact and disentangled unit-wise state representation outputted from transformer blocks, and produces its local action-value function. The proposed UNSR is used to boost the value decomposition with a multi-head attention mechanism for producing efficient credit assignment in the mixing network, providing an efficient reasoning path between the individual value function and joint value function. Experimental results demonstrate that our method achieves superior performance and data efficiency compared to solid baselines on the StarCraft II micromanagement challenge. Additional ablation experiments also help identify the key factors contributing to the performance of UNSR.</description><subject>Ablation</subject><subject>Decomposition</subject><subject>Multiagent systems</subject><subject>Representations</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>BENPR</sourceid><recordid>eNqNzcsKwjAQBdAgCIr2HwKuCzWp2m194UI3PpcSZFqm1ExNpq78eKP4Aa6Ge--B6Yi-0nocZ6lSPRF5XyVJoqYzNZnovnjNiTyjLeXZ1C3IJdzo3pBHRrLyiUaeLHJ8QQ8yZwbL-AR5YMMg99A48KEyX1yQkwuiBpz5ol1bM8Z5GUCgaMN-g_snbcE4G34ORbcwtYfodwditF4dF5u4cfRowfO1otbZMF1VNtazNNVZpv9Tbzr7UB0</recordid><startdate>20230512</startdate><enddate>20230512</enddate><creator>Zhao, Qingpeng</creator><creator>Zhu, Yuanyang</creator><creator>Liu, Zichuan</creator><creator>Wang, Zhi</creator><creator>Chen, Chunlin</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20230512</creationdate><title>Boosting Value Decomposition via Unit-Wise Attentive State Representation for Cooperative Multi-Agent Reinforcement Learning</title><author>Zhao, Qingpeng ; Zhu, Yuanyang ; Liu, Zichuan ; Wang, Zhi ; Chen, Chunlin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_28137443883</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Ablation</topic><topic>Decomposition</topic><topic>Multiagent systems</topic><topic>Representations</topic><toplevel>online_resources</toplevel><creatorcontrib>Zhao, Qingpeng</creatorcontrib><creatorcontrib>Zhu, Yuanyang</creatorcontrib><creatorcontrib>Liu, Zichuan</creatorcontrib><creatorcontrib>Wang, Zhi</creatorcontrib><creatorcontrib>Chen, Chunlin</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Zhao, Qingpeng</au><au>Zhu, Yuanyang</au><au>Liu, Zichuan</au><au>Wang, Zhi</au><au>Chen, Chunlin</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Boosting Value Decomposition via Unit-Wise Attentive State Representation for Cooperative Multi-Agent Reinforcement Learning</atitle><jtitle>arXiv.org</jtitle><date>2023-05-12</date><risdate>2023</risdate><eissn>2331-8422</eissn><abstract>In cooperative multi-agent reinforcement learning (MARL), the environmental stochasticity and uncertainties will increase exponentially when the number of agents increases, which puts hard pressure on how to come up with a compact latent representation from partial observation for boosting value decomposition. To tackle these issues, we propose a simple yet powerful method that alleviates partial observability and efficiently promotes coordination by introducing the UNit-wise attentive State Representation (UNSR). In UNSR, each agent learns a compact and disentangled unit-wise state representation outputted from transformer blocks, and produces its local action-value function. The proposed UNSR is used to boost the value decomposition with a multi-head attention mechanism for producing efficient credit assignment in the mixing network, providing an efficient reasoning path between the individual value function and joint value function. Experimental results demonstrate that our method achieves superior performance and data efficiency compared to solid baselines on the StarCraft II micromanagement challenge. Additional ablation experiments also help identify the key factors contributing to the performance of UNSR.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2023-05
issn 2331-8422
language eng
recordid cdi_proquest_journals_2813744388
source Free E- Journals
subjects Ablation
Decomposition
Multiagent systems
Representations
title Boosting Value Decomposition via Unit-Wise Attentive State Representation for Cooperative Multi-Agent Reinforcement Learning
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-11T13%3A16%3A30IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Boosting%20Value%20Decomposition%20via%20Unit-Wise%20Attentive%20State%20Representation%20for%20Cooperative%20Multi-Agent%20Reinforcement%20Learning&rft.jtitle=arXiv.org&rft.au=Zhao,%20Qingpeng&rft.date=2023-05-12&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2813744388%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2813744388&rft_id=info:pmid/&rfr_iscdi=true