Symmetries-enhanced Multi-Agent Reinforcement Learning
Not on the map. Why?
Abstract (arXiv)
Multi-agent reinforcement learning has emerged as a powerful framework for enabling agents to learn complex, coordinated behaviors but faces persistent challenges regarding its generalization, scalability and sample efficiency. Recent advancements have sought to alleviate those issues by embedding intrinsic symmetries of the systems in the policy. Yet, most dynamical systems exhibit little to no symmetries to exploit. This paper presents a novel framework for embedding extrinsic symmetries in multi-agent system dynamics that enables the use of symmetry-enhanced methods to address systems with insufficient intrinsic symmetries, expanding the scope of equivariant learning to a wide variety of MARL problems. Central to our framework is the Group Equivariant Graphormer, a group-modular architecture specifically designed for distributed swarming tasks. Extensive experiments on a swarm of symmetry-breaking quadrotors validate the effectiveness of our approach, showcasing its potential for improved generalization and zero-shot scalability. Our method achieves significant reductions in collision rates and enhances task success rates across a diverse range of scenarios and varying swarm sizes.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 55 of its 71 references that are in the map. First posted on arXiv 2 Jan 2025.
Builds on
55 of 71 references.
References · 55
- 2021E(n) Equivariant Graph Neural Networkscited by 951
- 2021The Surprising Effectiveness of PPO in Cooperative, Multi-Agent Gamescited by 903
- 2018Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point cloudscited by 735
- 2020SE(3)-Transformers: 3D Roto-Translation Equivariant Attention Networkscited by 602
- 2020Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixelscited by 573
- 2020Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learningcited by 475
- 2020Reinforcement Learning with Augmented Datacited by 452
- 2021Equivariant message passing for the prediction of tensorial properties and molecular spectracited by 438
- 2021Vector Neurons: A General Framework for SO(3)-Equivariant Networkscited by 304
- 2020Generalizing Convolutional Neural Networks for Equivariance to Lie Groups on Arbitrary Continuous Datacited by 248
- 2018Mean Field Multi-Agent Reinforcement Learningcited by 202
- 2020"Other-Play" for Zero-Shot Coordinationcited by 194
38 more
- 2018Graph Convolutional Reinforcement Learningcited by 148
- 2020MDP Homomorphic Networks: Group Symmetries in Reinforcement Learningcited by 133
- 2019Deep Coordination Graphscited by 96
- 2020Gauge Equivariant Mesh CNNs: Anisotropic convolutions on geometric graphscited by 90
- 2020Sample Factory: Egocentric 3D Control from Pixels at 100000 FPS with Asynchronous Reinforcement Learningcited by 84
- 2021Automatic Symmetry Discovery with Lie Algebra Convolutional Networkcited by 82
- 2023Attending to Graph Transformerscited by 81
- 2021Learning to Fly -- a Gym Environment with PyBullet Physics for Reinforcement Learning of Multi-agent Quadcopter Controlcited by 68
- 2020Plannable Approximations to MDP Homomorphisms: Equivariance under Actionscited by 61
- 2021Equivariant Q Learning in Spatial Action Spacescited by 61
- 2020Symmetry-Aware Actor-Critic for 3D Molecular Designcited by 55
- 2019Deep Reinforcement Learning meets Graph Neural Networks: exploring a routing optimization use casecited by 53
- 2022Sample Efficient Grasp Learning Using Equivariant Modelscited by 52
- 2019PIC: Permutation Invariant Critic for Multi-Agent Deep Reinforcement Learningcited by 46
- 2023Fast, Expressive SE(n) Equivariant Networks through Weight-Sharing in Position-Orientation Spacecited by 40
- 2022Edge Grasp Network: A Graph-Based SE(3)-invariant Approach to Grasp Detectioncited by 34
- 2022Scalable Multi-Agent Reinforcement Learning through Intelligent Information Aggregationcited by 34
- 2022The Surprising Effectiveness of Equivariant Models in Domains with Latent Symmetrycited by 32
- 2021Koopman Q-learning: Offline Reinforcement Learning via Symmetries of Dynamicscited by 27
- 2020Group Equivariant Deep Reinforcement Learningcited by 22
- 2021Enabling equivariance for arbitrary Lie groupscited by 22
- 2024Improving Equivariant Model Training via Constraint Relaxationcited by 22
- 2024Leveraging Symmetry in RL-based Legged Locomotion Controlcited by 19
- 2021Decentralized Control of Quadrotor Swarms with End-to-end Deep Reinforcement Learningcited by 18
- 2024Equivariant Reinforcement Learning under Partial Observabilitycited by 17
- 2020Policy learning in SE(3) action spacescited by 14
- 2022Graph Neural Networks for Multi-Robot Active Information Acquisitioncited by 12
- 2023Lie Group Decompositions for Equivariant Neural Networkscited by 12
- 2023Subequivariant Graph Reinforcement Learning in 3D Environmentscited by 10
- 2024Boosting Sample Efficiency and Generalization in Multi-agent Reinforcement Learning via Equivariancecited by 9
- 2018Policy Gradient With Value Function Approximation For Collective Multiagent Planningcited by 6
- 2023On discrete symmetries of robotics systems: A group-theoretic and data-driven analysiscited by 6
- 2022Communication-Efficient Actor-Critic Methods for Homogeneous Markov Gamescited by 5
- 2023SO(2)-Equivariant Downwash Models for Close Proximity Flightcited by 5
- 2023QuadSwarm: A Modular Multi-Quadrotor Simulator for Deep Reinforcement Learning with Direct Thrust Controlcited by 5
- 2020Goal-conditioned Batch Reinforcement Learning for Rotation Invariant Locomotioncited by 4
- 2022Equivariant Reinforcement Learning for Quadrotor UAVcited by 4
- 2022Exploiting Different Symmetries for Trajectory Tracking Control with Application to Quadrotorscited by 4
50 most cited of 55 All 55 (JSON)
Built on by
Citing papers per year: 2026: 1.
1 paper.