mlsanityData to 7 Oct 2026

MoE-Loco: Mixture of Experts for Multitask Locomotion

arXiv 2503.08564 ↗First posted 11 Mar 2025cs.RO cs.AIcited by 17

On the map: Mixture-of-experts → MoE in reinforcement learning and robotics

Abstract (arXiv)

We present MoE-Loco, a Mixture of Experts (MoE) framework for multitask locomotion for legged robots. Our method enables a single policy to handle diverse terrains, including bars, pits, stairs, slopes, and baffles, while supporting quadrupedal and bipedal gaits. Using MoE, we mitigate the gradient conflicts that typically arise in multitask reinforcement learning, improving both training efficiency and performance. Our experiments demonstrate that different experts naturally specialize in distinct locomotion behaviors, which can be leveraged for task migration and skill composition. We further validate our approach in both simulation and real-world deployment, showcasing its robustness and adaptability.

Key facts. Cited by 17 papers in the map (data through 7 Oct 2026). Placed in MoE in reinforcement learning and robotics, one of 12 directions in Mixture-of-experts. Builds on 39 of its 54 references that are in the map. First posted on arXiv 11 Mar 2025.

Builds on

39 of 54 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024MENTOR: Mixture-of-Experts Network with Task-Oriented Perturbation for Visual Reinforcement Learning19 Oct
  2. 2024Acquiring Diverse Skills using Curriculum Reinforcement Learning with Mixture of Experts11 Mar
  3. 2024Mixtures of Experts Unlock Parameter Scaling for Deep RL13 Feb

Other references · 36

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
  3. 2024Mixtral of Expertscited by 1,785
  4. 2020GShard: Scaling Giant Models with Conditional Computation and Automatic Shardingcited by 1,601
  5. 2020Gradient Surgery for Multi-Task Learningcited by 1,210
  6. 2021Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learningcited by 882
  7. 2020Learning Quadrupedal Locomotion over Challenging Terraincited by 494
  8. 2016RL2: Fast Reinforcement Learning via Slow Reinforcement Learningcited by 490
  9. 2021Conflict-Averse Gradient Descent for Multi-task Learningcited by 441
  10. 2019Learning agile and dynamic motor skills for legged robotscited by 402
  11. 2021Multi-Task Reinforcement Learning with Context-based Representationscited by 169
  12. 2016Recurrent Neural Network for Text Classification with Multi-Task Learningcited by 144
24 more
  1. 2022Rapid Locomotion via Reinforcement Learningcited by 130
  2. 2023Extreme Parkour with Legged Robotscited by 120
  3. 2023Robot Parkour Learningcited by 115
  4. 2023Mixture of Cluster-conditional LoRA Experts for Vision-language Instruction Tuningcited by 106
  5. 2023ANYmal Parkour: Learning Agile Navigation for Quadrupedal Robotscited by 105
  6. 2023GNFactor: Multi-Task Real Robot Learning with Generalizable Neural Feature Fieldscited by 99
  7. 2015Distributed Gaussian Processescited by 84
  8. 2024LLaVA-MoLE: Sparse Mixture of LoRA Experts for Mitigating Data Conflicts in Instruction Finetuning MLLMscited by 84
  9. 2020Multi-expert learning of adaptive legged locomotioncited by 62
  10. 2024Agile But Safe: Learning Collision-Free High-Speed Legged Locomotioncited by 53
  11. 2024Learning Humanoid Locomotion with Perceptive Internal Modelcited by 38
  12. 2016Learning to Push by Grasping: Using multiple tasks for effective learningcited by 36
  13. 2024M3oE: Multi-Domain Multi-Task Mixture-of Experts Recommendation Frameworkcited by 32
  14. 2024Leveraging Symmetry in RL-based Legged Locomotion Controlcited by 19
  15. 2024PIE: Parkour with Implicit-Explicit Learning Framework for Legged Robotscited by 17
  16. 2023ManyQuadrupeds: Learning a Single Locomotion Policy for Diverse Quadruped Robotscited by 15
  17. 2023Improving Generalization in Visual Reinforcement Learning via Conflict-aware Gradient Agreement Augmentationcited by 13
  18. 2022AdaMCT: Adaptive Mixture of CNN-Transformer for Sequential Recommendationcited by 11
  19. 2023Learning Agile Bipedal Motions on a Quadrupedal Robotcited by 10
  20. 2024Quadruped robot traversing 3D complex environments with limited perceptioncited by 8
  21. 2024Robust Robot Walker: Learning Agile Locomotion over Tiny Trapscited by 7
  22. 2024SARO: Space-Aware Robot System for Terrain Crossing via Vision-Language Modelcited by 4
  23. 2023Learning a Single Policy for Diverse Behaviors on a Quadrupedal Robot using Scalable Motion Imitationcited by 3
  24. 2023MTAC: Hierarchical Reinforcement Learning-based Multi-gait Terrain-adaptive Quadruped Controllercited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 6; 2026: 11.

17 papers.

Sort:

By area:

  1. 2026Hiking in the Wild: A Scalable Perceptive Parkour Framework for Humanoidscited by 11
  2. 2025MoE-DP: An MoE-Enhanced Diffusion Policy for Robust Long-Horizon Robotic Manipulation with Skill Decomposition and Failure Recoverycited by 9
  3. 2025Dynamic Mixture of Progressive Parameter-Efficient Expert Library for Lifelong Robot Learningcited by 8
  4. 2025Flexible Multitask Learning with Factorized Diffusion Policycited by 5
  5. 2026MoE Adapter for Large Audio Language Models: Sparsity, Disentanglement, and Gradient-Conflict-Freecited by 4
  6. 2025Generative Large-Scale Pre-trained Models for Automated Ad Bidding Optimizationcited by 3
  7. 2026TTT-Parkour: Rapid Test-Time Training for Perceptive Robot Parkourcited by 3
  8. 2026SPHERE: Mitigating the Loss of Spectral Plasticity in Mixture-of-Experts for Deep Reinforcement Learningcited by 2
  9. 2026SARM2: Multi-Task Stage Aware Reward Modeling for Self Improving Robotic Manipulationcited by 2
  10. 2025Flexible Locomotion Learning with Diffusion Model Predictive Controlcited by 1
  11. 2026LiMoDE: Rethinking Lifelong Robot Manipulation from a Mixture-of-Dynamic-Experts Perspectivecited by 1
  12. 2025MoE-CE: Enhancing Generalization for Deep Learning based Channel Estimation via a Mixture-of-Experts Frameworkcited by 0
5 more
  1. 2026CoRe-MoE: Contrastive Reweighted Mixture of Experts for Multi-Terrain Humanoid Locomotion with Gait Adaptationcited by 0
  2. 2026Y-BotFrame: An Extensible Embodied Agent Framework for Quadruped Robot Assistantscited by 0
  3. 2026CTS-MoE: Implicit Terrain Adaptation via Mixture-of-Experts for Perceptive Locomotioncited by 0
  4. 2026CoRDE: Concept-Prior Routed Diffusion Experts for Structural Generalization in Robot Manipulationcited by 0
  5. 2026Self-Adaptive Learning and Model Predictive Control for Tracking Unknown Dynamics with No Regretcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Multi-Loco: Unifying Multi-Embodiment Legged Locomotion via Reinforcement Learning Augmented Diffusion
  2. 2025GRoQ-LoCO: Generalist and Robot-agnostic Quadruped Locomotion Control using Offline Datasets
  3. 2026LocoVLM: Grounding Vision and Language for Adapting Versatile Legged Locomotion Policies
  4. 2023LocoMuJoCo: A Comprehensive Imitation Learning Benchmark for Locomotion
  5. 2021Specializing Versatile Skill Libraries using Local Mixture of Experts
  6. 2025SkillBlender: Towards Versatile Humanoid Whole-Body Loco-Manipulation via Skill Blending
  7. 2024HYPERmotion: Learning Hybrid Behavior Planning for Autonomous Loco-manipulation
  8. 2025STRIDE: Automating Reward Design, Deep Reinforcement Learning Training and Feedback Optimization in Humanoid Robotics Locomotion