AI Paper Atlas

Research map / Multi-agent systems and games / Cooperative multi-agent RL

Scalable Multi-Agent Reinforcement Learning through Intelligent Information Aggregation

Siddharth Nayak, Kenneth Choi, Wenqi Ding, Sydney Dolan, Karthik Gopalakrishnan, Hamsa Balakrishnan

ICML 2023 · Multi-agent systems and games › Cooperative multi-agent RL · approach: MARL benchmarks and grouping · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Aggregate local agent and entity information with an attention-based graph networkMAPPO for multi-agent navigation · Target, 3, 7, and 10 agentsReward: 173.13 -> 205.24; 327.39 -> 399.01; 366.81 -> 429.14Improved
Use the local-information graph aggregation architecture instead of RMAPPO with global informationMAPPO for multi-agent navigation · Target, 3, 7, and 10 agentsSuccess rate: 96 -> 100; 88 -> 100; 79 -> 100Improved
Add the graph information aggregation module to the centralized critic rather than concatenating agent states (ablation)InforMARL · Target, 3-agent trainingTraining performanceNo clear effect
Add the graph information aggregation module to the centralized critic rather than concatenating agent states (ablation)InforMARL · Target, variable numbers of entitiesCritic parameter countImproved
Increase sensing radius from small values (ρ=0.1 or ρ=0.2) to larger values (ρ=0.5, ρ=1, ρ=2, or ρ=5) (ablation)InforMARL · Target taskRewardImproved
Increase sensing radius from ρ=0.5 to ρ=5 (ablation)InforMARL · Target taskRewardNo clear effect
Use one-hot encoding instead of an embedding layer for entity types (ablation)InforMARL · Multi-agent navigationPerformanceNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy