mlsanityData to 14 Oct 2026

DriveMoE: Mixture-of-Experts for Vision-Language-Action Model in End-to-End Autonomous Driving

arXiv 2505.16278 ↗First posted 22 May 2025cs.CV cs.AI cs.ROcited by 86

On the map: Vision-language-action models → VLA for autonomous driving

also in 2 more
Abstract (arXiv)

End-to-end autonomous driving (E2E-AD) demands effective processing of multi-view sensory data and robust handling of diverse and complex driving scenarios, particularly rare maneuvers such as aggressive turns. Recent success of Mixture-of-Experts (MoE) architecture in Large Language Models (LLMs) demonstrates that specialization of parameters enables strong scalability. In this work, we propose DriveMoE, a novel MoE-based E2E-AD framework, with a Scene-Specialized Vision MoE and a Skill-Specialized Action MoE. DriveMoE is built upon our π0 Vision-Language-Action (VLA) baseline (originally from the embodied AI field), called Drive-π0. Specifically, we add Vision MoE to Drive-π0 by training a router to select relevant cameras according to the driving context dynamically. This design mirrors human driving cognition, where drivers selectively attend to crucial visual cues rather than exhaustively processing all visual information. In addition, we add Action MoE by training another router to activate specialized expert modules for different driving behaviors. Through explicit behavioral specialization, DriveMoE is able to handle diverse scenarios without suffering from modes averaging like existing models. In Bench2Drive closed-loop evaluation experiments, DriveMoE achieves state-of-the-art (SOTA) performance, demonstrating the effectiveness of combining vision and action MoE in autonomous driving tasks. We will release our code and models of DriveMoE and Drive-π0.

Key facts. Cited by 86 papers in the map (data through 14 Oct 2026). Placed in VLA for autonomous driving, one of 13 directions in Vision-language-action models. Builds on 48 of its 70 references that are in the map. First posted on arXiv 22 May 2025.

Builds on

48 of 70 references.

References · 48

  1. 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,137
  2. 2022Flow Matching for Generative Modelingcited by 5,702
  3. 2023Qwen Technical Reportcited by 3,344
  4. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  5. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,838
  6. 2017CARLA: An Open Urban Driving Simulatorcited by 2,449
  7. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  8. 2024Mixtral of Expertscited by 1,823
  9. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,761
  10. 2022Planning-oriented Autonomous Drivingcited by 1,087
  11. 2024DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Modelscited by 867
  12. 2024PaliGemma: A versatile 3B VLM for transfercited by 696
36 more
  1. 2023VAD: Vectorized Scene Representation for Efficient Autonomous Drivingcited by 622
  2. 2023DriveLM: Driving with Graph Visual Question Answeringcited by 536
  3. 2022TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Drivingcited by 485
  4. 2023DriveGPT4: Interpretable End-to-end Autonomous Driving via Large Language Modelcited by 461
  5. 2023LMDrive: Closed-Loop End-to-End Driving with Large Language Modelscited by 305
  6. 2022Rectified Flow: A Marginal Preserving Approach to Optimal Transportcited by 283
  7. 2024Bench2Drive: Towards Multi-Ability Benchmarking of Closed-Loop End-To-End Autonomous Drivingcited by 252
  8. 2022Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baselinecited by 248
  9. 2024A Survey on Mixture of Experts in Large Language Modelscited by 237
  10. 2023DriveMLM: Aligning Multi-Modal Large Language Models with Behavioral Planning States for Autonomous Drivingcited by 233
  11. 2024GenAD: Generative End-to-End Autonomous Drivingcited by 225
  12. 2023Rethinking the Open-Loop Evaluation of End-to-End Autonomous Driving in nuScenescited by 191
  13. 2023Efficient Large Language Models: A Surveycited by 186
  14. 2025ORION: A Holistic End-to-End Autonomous Driving Framework by Vision-Language Instructed Action Generationcited by 178
  15. 2023Think Twice before Driving: Towards Scalable Decoders for End-to-End Autonomous Drivingcited by 169
  16. 2025End-to-End Driving with Online Trajectory Evaluation via BEV World Modelcited by 151
  17. 2025SimLingo: Vision-Only Closed-Loop Autonomous Driving with Language-Action Alignmentcited by 150
  18. 2023DriveAdapter: Breaking the Coupling Barrier of Perception and Planning in End-to-End Autonomous Drivingcited by 144
  19. 2023LLM4Drive: A Survey of Large Language Models for Autonomous Drivingcited by 136
  20. 2025DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Drivingcited by 126
  21. 2024LLaMA-MoE: Building Mixture-of-Experts from LLaMA with Continual Pre-trainingcited by 123
  22. 2022Delving into the Devils of Bird's-eye-view Perception: A Review, Evaluation and Recipecited by 122
  23. 2022HDGT: Heterogeneous Driving Graph Transformer for Multi-Agent Trajectory Prediction via Scene Encodingcited by 110
  24. 2023Hidden Biases of End-to-End Driving Modelscited by 96
  25. 2025Don't Shake the Wheel: Momentum-Aware Planning in End-to-End Autonomous Drivingcited by 84
  26. 2023GPT4Point: A Unified Framework for Point-Language Understanding and Generationcited by 71
  27. 2025Interleave-VLA: Enhancing Robot Manipulation with Interleaved Image-Text Instructionscited by 51
  28. 2025Distilling Multi-modal Large Language Models for Autonomous Drivingcited by 46
  29. 2024Continuously Learning, Adapting, and Improving: A Dual-Process Approach to Autonomous Drivingcited by 41
  30. 2025DiffAD: A Unified Diffusion Modeling Approach for Autonomous Drivingcited by 26
  31. 2024MENTOR: Mixture-of-Experts Network with Task-Oriented Perturbation for Visual Reinforcement Learningcited by 23
  32. 2020IDE-Net: Interactive Driving Event and Pattern Extraction from Human Datacited by 19
  33. 2024Bench2Drive-R: Turning Real World Data into Reactive Closed-Loop Autonomous Driving Benchmark by Generative Modelcited by 14
  34. 2024AMP: Autoregressive Motion Prediction Revisited with Next Token Prediction for Autonomous Drivingcited by 13
  35. 2024FlatFusion: Delving into Details of Sparse Transformer-based Camera-LiDAR Fusion for Autonomous Drivingcited by 9
  36. 2024EgoFSD: Ego-Centric Fully Sparse Paradigm with Uncertainty Denoising and Iterative Refinement for Efficient End-to-End Self-Drivingcited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 24; 2026: 62.

86 papers.

Sort:

By area:

  1. 2025DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Drivingcited by 134
  2. 2025A Survey on Vision-Language-Action Models for Autonomous Drivingcited by 72
  3. 2025Pure Vision Language Action (VLA) Models: A Comprehensive Surveycited by 49
  4. 2025ReSim: Reliable World Simulation for Autonomous Drivingcited by 44
  5. 2026DriveWorld-VLA: Unified Latent-Space World Modeling with Vision-Language-Action for Autonomous Drivingcited by 37
  6. 2025AutoDrive-R2: Incentivizing Reasoning and Self-Reflection Capacity for VLA Model in Autonomous Drivingcited by 34
  7. 2025Vision-Language-Action Models for Autonomous Driving: Past, Present, and Futurecited by 33
  8. 2025SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Drivingcited by 28
  9. 2025MindDrive: A Vision-Language-Action Model for Autonomous Driving via Online Reinforcement Learningcited by 25
  10. 2026DriveVA: Video Action Models are Zero-Shot Driverscited by 23
  11. 2026PALM: Progress-Aware Policy Learning via Affordance Reasoning for Long-Horizon Robotic Manipulationcited by 21
  12. 2026UniDriveVLA: Unifying Understanding, Perception, and Action Planning for Autonomous Drivingcited by 19
38 more
  1. 2026DriveDreamer-Policy: A Geometry-Grounded World-Action Model for Unified Generation and Planningcited by 17
  2. 2026AtomicVLA: Unlocking the Potential of Atomic Skill Learning in Robotscited by 16
  3. 2026EponaV2: Driving World Model with Comprehensive Future Reasoningcited by 15
  4. 2025Reasoning-VLA: A Fast and General Vision-Language-Action Reasoning Model for Autonomous Drivingcited by 14
  5. 2026DriveWAM: Video Generative Priors Enable Scalable World-Action Modeling for Autonomous Drivingcited by 14
  6. 2026DVGT-2: Vision-Geometry-Action Model for Autonomous Driving at Scalecited by 13
  7. 2025Expertise need not monopolize: Action-Specialized Mixture of Experts for Vision-Language-Action Learningcited by 12
  8. 2026CoWorld-VLA: Thinking in a Multi-Expert World Model for Autonomous Drivingcited by 12
  9. 2026SteerVLA: Steering Vision-Language-Action Models in Long-Tail Driving Scenarioscited by 11
  10. 2026DynVLA: Learning World Dynamics for Action Reasoning in Autonomous Drivingcited by 11
  11. 2026Learning Vision-Language-Action World Models for Autonomous Drivingcited by 11
  12. 2025UniUGP: Unifying Understanding, Generation, and Planing For End-to-end Autonomous Drivingcited by 10
  13. 2025WAM-Diff: A Masked Diffusion VLA Framework with MoE and Online Reinforcement Learning for Autonomous Drivingcited by 10
  14. 2026FLARE: Learning Future-Aware Latent Representations from Vision-Language Models for Autonomous Drivingcited by 10
  15. 2026Bench2Drive-VL: Benchmarks for Closed-Loop Autonomous Driving with Vision-Language Modelscited by 10
  16. 2026GuidedVLA: Specifying Task-Relevant Factors via Plug-and-Play Action Attention Specializationcited by 10
  17. 2026DriveMamba: Task-Centric Scalable State Space Model for Efficient End-to-End Autonomous Drivingcited by 9
  18. 2026Unifying Language-Action Understanding and Generation for Autonomous Drivingcited by 9
  19. 2026CORAL: Scalable Multi-Task Robot Learning via LoRA Expertscited by 9
  20. 2026AutoMoT: A Unified Vision-Language-Action Model with Asynchronous Mixture-of-Transformers for End-to-End Autonomous Drivingcited by 9
  21. 2026Learning from Mistakes: Post-Training for Driving VLA with Takeover Datacited by 8
  22. 2026World Engine: Towards the Era of Post-Training for Autonomous Drivingcited by 8
  23. 2025UniMM-V2X: MoE-Enhanced Multi-Level Fusion for End-to-End Cooperative Autonomous Drivingcited by 6
  24. 2026Efficient-LVSM: Faster, Cheaper, and Better Large View Synthesis Model via Decoupled Co-Refinement Attentioncited by 6
  25. 2026Vega: Learning to Drive with Natural Language Instructionscited by 6
  26. 2025A Knowledge-Driven Diffusion Policy for End-to-End Autonomous Driving Based on Expert Routingcited by 5
  27. 2025Spatial Retrieval Augmented Autonomous Drivingcited by 5
  28. 2025MonitorVLM:A Vision Language Framework for Safety Violation Detection in Mining Operationscited by 4
  29. 2025DriveVGGT: Calibration-Constrained Visual Geometry Transformers for Multi-Camera Autonomous Drivingcited by 4
  30. 2026NaviDriveVLM: Decoupling High-Level Reasoning and Motion Planning for Autonomous Drivingcited by 4
  31. 2026SAMoE-VLA: A Scene Adaptive Mixture-of-Experts Vision-Language-Action Model for Autonomous Drivingcited by 4
  32. 2026ChainFlow-VLA: Causal Flow Planning with Vision-Language Modelscited by 4
  33. 2025TopoStreamer: Temporal Lane Segment Topology Reasoning in Autonomous Drivingcited by 3
  34. 2025MoSE: Skill-by-Skill Mixture-of-Experts Learning for Embodied Autonomous Machinescited by 3
  35. 2025Large Foundation Models for Trajectory Prediction in Autonomous Driving: A Comprehensive Surveycited by 3
  36. 2026Bench2Drive-Robust: Benchmarking Closed-Loop Autonomous Driving under Deployment Perturbationscited by 3
  37. 2026VLGA: Vision-Language-Geometry-Action Models for Autonomous Drivingcited by 3
  38. 2025SafeCoop: Unravelling Full Stack Safety in Agentic Collaborative Drivingcited by 2

50 most cited of 86 All 86 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025OpenDriveVLA: Towards End-to-end Autonomous Driving with Large Vision Language Action Model
  2. 2025ExpertAD: Enhancing Autonomous Driving Systems with Mixture of Experts
  3. 2025Sce2DriveX: A Generalized MLLM Framework for Scene-to-Drive Learning
  4. 2025InsightDrive: Insight Scene Representation for End-to-End Autonomous Driving
  5. 2025DriveAction: A Benchmark for Exploring Human-like Driving Decisions in VLA Models
  6. 2026DriveWLA: Unifying Future Dynamics and Scene Semantics through World-Language-Action Modeling for Autonomous Driving
  7. 2025AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning
  8. 2026DriveZero: End-to-End Driving Beyond Human Demonstrations