This repo contains the paper list for Foundation Models for Embodied Navigation: A Survey.
The remarkable advancements of embodied navigation, the process of physically situated agents perceiving and reasoning through egocentric observations to reach target locations, have recently been reshaped by the emergence of foundation models. Departing from traditional, task-specific policies trained from scratch on limited datasets, this new paradigm leverages the reasoning and multimodal capabilities of Large Language Models (LLMs), Vision-Language Models (VLMs), and Video Generation Models (VGMs), achieving superior generalization and flexible decision-making in unseen environments. For the first time, this survey systematically reviews the landscape of foundation models for embodied navigation, focusing on systems where these models play a central role in perception, long-horizon memory management, and action generation. The survey aims to categorize and interpret existing embodied navigation research through the lens of design paradigms, data sources, and training strategies. Through this analysis, we offer a synthesized outlook on the evolution of navigation brains, highlight the bottlenecks of dataset bias, and contribute guidance for future research, hoping to bring the field closer to robust, general-purpose embodied intelligence.
Overview of embodied navigation systems.
- Foundation Models for Embodied Navigation: A Survey
This paper list is organized following the structure of the survey. Papers are grouped by the corresponding sections and subsections in the survey, so readers can directly trace each reference back to the topic where it is discussed.
- This work provides a comprehensive and systematic overview of the current landscape of foundation models within the domain of embodied navigation. We explicitly define foundation models for embodied navigation as those that leverage pre-trained language or multimodal foundation models as the core backbone of their decision-making pipeline.
- Our scope is defined by two key exclusions: (1) we omit methods that treat LLMs/VLMs merely as external, modular tools for auxiliary tasks (such as captioning or object detection) without integrating them into the central control logic; and (2) we exclude the field of autonomous driving, as it represents a mature, distinct domain with environmental constraints and safety requirements that differ significantly from the general embodied navigation tasks focused on in this survey.
- We primarily survey papers published in top-tier computer science and robotics conferences or journals, specifically those included in CSRankings. We also manually pick related and potentially high-impact papers from arXiv.
- We mainly survey papers published after the year of 2023, since the innovation of AI is going fast with old knowledge and methods being overturned frequently.
- Vla-rl: Towards masterful and general robotic manipulation with scalable reinforcement learning. [arXiv'25] [Paper]
- Precise and dexterous robotic manipulation via human-in-the-loop reinforcement learning. [Science Robotics'25] [Paper]
-
$\pi_0$ : A Vision-Language-Action Flow Model for General Robot Control. [RSS'25] [Paper] -
$\pi_{0.5}$ : a Vision-Language-Action Model with Open-World Generalization. [arXiv'25] [Paper] -
$\pi^*_{0.6}$ : a VLA That Learns From Experience. [arXiv'25] [Paper]
- Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments. [CVPR'18] [Paper]
- Room-across-room: Multilingual vision-and-language navigation with dense spatiotemporal grounding. [EMNLP'20] [Paper]
- Beyond the nav-graph: Vision-and-language navigation in continuous environments. [ECCV'20] (https://link.springer.com/chapter/10.1007/978-3-030-58604-1_7)
- Rethinking the embodied gap in vision-and-language navigation: A holistic study of physical and visual disparities. [ICCV'25] [Paper]
- On evaluation of embodied navigation agents. [arXiv'18] [Paper]
- ObjectNav Revisited: On Evaluation of Embodied Agents Navigating to Objects. [arXiv'20] [Paper]
- Hm3d-ovon: A dataset and benchmark for open-vocabulary object goal navigation. [IROS'24] [Paper]
- FloorPlan-VLN: A New Paradigm for Floor Plan Guided Vision-Language Navigation. [arXiv'26] [Paper]
- The surprising effectiveness of visual odometry techniques for embodied pointgoal navigation. [ICCV'21] [Paper]
- Neural topological slam for visual navigation. [CVPR'20] [Paper]
- Instance-Specific Image Goal Navigation: Training Embodied Agents to Find Object Instances. [arXiv'22] [Paper]
- Towards distraction-robust active visual tracking. [ICML'21] [Paper]
- Empowering embodied visual tracking with visual foundation models and offline rl. [ECCV'24] [Paper]
- Trackvla: Embodied visual tracking in the wild. [arXiv'25] [Paper]
- Socially aware motion planning with deep reinforcement learning. [IROS'17] [Paper]
- Autonomous human--robot proxemics: socially aware navigation based on interaction potential. [Autonomous Robots'17] [Paper]
- Vlm-social-nav: Socially aware robot navigation through scoring using vision-language models. [RA-L'24] [Paper]
- Socially compliant navigation dataset (scand): A large-scale dataset of demonstrations for social navigation. [RA-L'22] [Paper]
- Multi-agent embodied visual semantic navigation with scene prior knowledge. [RA-L'22] [Paper]
- Qwen3 technical report. [arXiv'25] [Paper]
- Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks. [CVPR'24] [Paper]
- Wan: Open and advanced large-scale video generative models. [arXiv'25] [Paper]
- V-jepa 2: Self-supervised video models enable understanding, prediction and planning. [arXiv'25] [Paper]
- Genie: Generative interactive environments. [ICML'24] [Paper]
- NaVILA: Legged Robot Vision-Language-Action Model for Navigation. [RSS'25] [Paper]
- InternVLA-N1: An Open Dual-System Navigation Foundation Model with Learned Latent Plans. [Preprint'25] [Paper]
- Ground slow, move fast: A dual-system foundation model for generalizable vision-and-language navigation. [ICLR'26] [Paper]
- Janusvln: Decoupling semantics and spatiality with dual implicit memory for vision-language navigation. [ICLR'26] [Paper]
- VLN-R1: Vision-Language Navigation via Reinforcement Fine-Tuning. [arXiv'25] [Paper]
- Embodied navigation foundation model. [ICLR'26] [Paper]
- Instructnav: Zero-shot system for generic instruction navigation in unexplored environment. [CoRL'24] [Paper]
- Wmnav: Integrating vision-language models into world models for object goal navigation. [IROS'25] [Paper]
- VLA-AN: An Efficient and Onboard Vision-Language-Action Framework for Aerial Navigation in Complex Environments. [arXiv'25] [Paper]
- Nav-r1: Reasoning and navigation in embodied scenes. [arXiv'25] [Paper]
- Na Vid-4D: Unleashing Spatial Intelligence in Egocentric RGB-D Videos for Vision-and-Language Navigation. [ICRA'25] [Paper]
- FOM-Nav: Frontier-Object Maps for Object Goal Navigation. [arXiv'25] [Paper]
- Cognav: Cognitive process modeling for object goal navigation with llms. [ICCV'25] [Paper]
- RoboTron-Nav: A Unified Framework for Embodied Navigation Integrating Perception, Planning, and Prediction. [ICCV'25] [Paper]
- NavForesee: A Unified Vision-Language World Model for Hierarchical Planning and Dual-Horizon Navigation Prediction. [arXiv'25] [Paper]
- AstraNav-Memory: Contexts Compression for Long Memory. [arXiv'25] [Paper]
- Uni-navid: A video-based vision-language-action model for unifying embodied navigation tasks. [RSS'25] [Paper]
- Navid: Video-based vlm plans the next step for vision-and-language navigation. [RSS'24] [Paper]
- VLingNav: Embodied Navigation with Adaptive Reasoning and Visual-Assisted Linguistic Memory. [arXiv'26] [Paper]
- Nav-$R^2$ Dual-Relation Reasoning for Generalizable Open-Vocabulary Object-Goal Navigation. [arXiv'25] [Paper]
- Streamvln: Streaming vision-and-language navigation via slowfast context modeling. [arXiv'25] [Paper]
- Hydra-Nav: Object Navigation via Adaptive Dual-Process Reasoning. [arXiv'26] [Paper]
- OmniNav: A Unified Framework for Prospective Exploration and Visual-Language Navigation. [arXiv'25] [Paper]
- Octonav: Towards generalist embodied navigation. [arXiv'25] [Paper]
- Mapnav: A novel memory representation via annotated semantic maps for vlm-based vision-and-language navigation. [ACL'25] [Paper]
- Mem2ego: Empowering vision-language models with global-to-ego memory for long-horizon embodied navigation. [arXiv'25] [Paper]
- GSMem: 3D Gaussian Splatting as Persistent Spatial Memory for Zero-Shot Embodied Exploration and Reasoning. [arXiv'26] [Paper]
- 3d gaussian splatting for real-time radiance field rendering. [SIGGRAPH'23] [Paper]
- Sparse Video Generation Propels Real-World Beyond-the-View Vision-Language Navigation. [arXiv'26] [Paper]
- ABot-N0: Technical Report on the VLA Foundation Model for Versatile Embodied Navigation. [arXiv'26] [Paper]
- Aux-Think: Exploring Reasoning Strategies for Data-Efficient Vision-Language Navigation. [NeurIPS'25] [Paper]
- FantasyVLN: Unified Multimodal Chain-of-Thought Reasoning for Vision-Language Navigation. [arXiv'26] [Paper]
- AutoFly: Vision-Language-Action Model for UAV Autonomous Navigation in the Wild. [ICLR'26] [Paper]
- Urbanvla: A vision-language-action model for urban micromobility. [arXiv'25] [Paper]
- Socialnav: Training human-inspired foundation model for socially-aware embodied navigation. [arXiv'25] [Paper]
- MapDream: Task-Driven Map Learning for Vision-Language Navigation. [arXiv'26] [Paper]
- Matterport3d: Learning from rgb-d data in indoor environments. [arXiv'17] [Paper]
- Habitat-matterport 3d dataset (hm3d): 1000 large-scale 3d environments for embodied ai. [arXiv'21] [Paper]
- Gibson env: Real-world perception for embodied agents. [CVPR'18] [Paper]
- The replica dataset: A digital replica of indoor spaces. [arXiv'19] [Paper]
- Habitat synthetic scenes dataset (hssd-200): An analysis of 3d scene scale and realism tradeoffs for objectgoal navigation. [CVPR'24] [Paper]
- 3d-front: 3d furnished rooms with layouts and semantics. [ICCV'21] [Paper]
- Sensaturban: Learning semantics from urban-scale photogrammetric point clouds. [IJCV'22] [Paper]
- Aerialvln: Vision-and-language navigation for uavs. [ICCV'23] [Paper]
- Towards realistic uav vision-language navigation: Platform, benchmark, and methodology. [arXiv'24] [Paper]
- OpenFly: A comprehensive platform for aerial vision-language navigation. [ICLR'26] [Paper]
- Citynav: A large-scale dataset for real-world aerial navigation. [ICCV'25] [Paper]
- Scaling data generation in vision-and-language navigation. [ICCV'23] [Paper]
- Learning to navigate unseen environments: Back translation with environmental dropout. [ACL'19] [Paper]
- nuscenes: A multimodal dataset for autonomous driving. [CVPR'20] [Paper]
- Sacson: Scalable autonomous control for social navigation. [RA-L'23] [Paper]
- Rapid Exploration for Open-World Navigation with Latent Goal Models. [CoRL'21] [Paper]
- Grounding image matching in 3d with mast3r. [ECCV'24] [Paper]
- Roomtour3d: Geometry-aware video-instruction tuning for embodied navigation. [CVPR'25] [Paper]
- Lelan: Learning a language-conditioned navigation policy from in-the-wild videos. [CoRL'24] [Paper]
- Citywalker: Learning embodied urban navigation from web-scale videos. [CVPR'25] [Paper]
- Sekai: A video dataset towards world exploration. [arXiv'25] [Paper]
- Scanqa: 3d question answering for spatial scene understanding. [CVPR'22] [Paper]
- LLaVA-Video: Video Instruction Tuning With Synthetic Data [TMLR'25] [Paper]
- Multimodal c4: An open, billion-scale corpus of images interleaved with text. [NeurIPS'23] [Paper]
- RynnBrain: Open Embodied Foundation Models. [arXiv'26] [Paper]
- OpenVLN: Open-world Aerial Vision-Language Navigation. [arXiv'25] [Paper]
- OmniVLA: An omni-modal vision-language-action model for robot navigation. [arXiv'25] [Paper]
- X-Nav: Learning End-to-End Cross-Embodiment Navigation for Mobile Robots. [RA-L'25] [Paper]
- TIC-VLA: A Think-in-Control Vision-Language-Action Model for Robot Navigation in Dynamic Environments. [arXiv'26] [Paper]
- Quar-vla: Vision-language-action model for quadruped robots. [ECCV'24] [Paper]
- SkyVLN: Vision-and-Language Navigation and NMPC Control for UAVs in Urban Environments. [IROS'25] [Paper]
- LongFly: Long-Horizon UAV Vision-and-Language Navigation with Spatiotemporal Context Integration. [arXiv'25] [Paper]
- Asyncvla: An asynchronous vla for fast and robust navigation on the edge. [arXiv'26] [Paper]
- AirUniNav: Unified Vision-Language Navigation for UAVs in Indoor and Outdoor Scenes. [TechRxiv'26] [Paper]
- EfficientNav: Towards On-Device Object-Goal Navigation with Navigation Map Caching and Retrieval. [NeurIPS'25] [Paper]
- Reverie: Remote embodied visual referring expression in real indoor environments. [CVPR'20] [Paper]
- Vision-and-dialog navigation. [CoRL'20] [Paper]
- Soon: Scenario oriented object navigation with graph-based exploration. [CVPR'21] [Paper]
- Dynam3D: Dynamic layered 3D tokens empower VLM for vision-and-language navigation. [NeurIPS '25] [Paper]
- Vision and language navigation in the real world via online visual language mapping. [arXiv'23] [Paper]
- Mobility vla: Multimodal instruction navigation with long-context vlms and topological graphs. [CORL'24] [Paper]
- Bridging the Indoor-Outdoor Gap: Vision-Centric Instruction-Guided Embodied Navigation for the Last Meters. [arXiv'26] [Paper]
- Aerial vision-and-dialog navigation. [ACL'23] [Paper]
- Towards long-horizon vision-language navigation: Platform, benchmark and method. [CVPR'25] [Paper]
- Goat-bench: A benchmark for multi-modal lifelong navigation. [CVPR'24] [Paper]
- Robothor: An open simulation-to-real embodied ai platform. [CVPR'20] [Paper]
- Find what you want: Learning demand-conditioned object attribute space for demand-driven navigation. [NeurIPS'23] [Paper]
- LCLA: Language-Conditioned Latent Alignment for Vision-Language Navigation. [arXiv'26] [Paper]
- What You See is What You Reach: Towards Spatial Navigation with High-Level Human Instructions. [AAAI'26] [Paper]
- Tag map: A text-based map for spatial reasoning and navigation with large language models. [arXiv'24] [Paper]
- Metaurban: An embodied ai simulation platform for urban micromobility. [ICLR'25] [Paper]
- Navsim: Data-driven non-reactive autonomous vehicle simulation and benchmarking. [NeurIPS'24] [Paper]
- Semantickitti: A dataset for semantic scene understanding of lidar sequences. [ICCV'19] [Paper]
- Openeqa: Embodied question answering in the era of foundation models. [CVPR'24] [Paper]
- Embodied question answering in photorealistic environments with point cloud perception. [CVPR'19] [Paper]
- Sqa3d: Situated question answering in 3d scenes. [ICLR'23] [Paper]
- 3d-llm: Injecting the 3d world into large language models. [NeurIPS'23] [Paper]
- Habitat 3.0: A co-habitat for humans, avatars and robots. [ICLR'24] [Paper]
- Socnavbench: A grounded simulation testing framework for evaluating social navigation. [THRI'22] [Paper]
- Olivia-nav: An online lifelong vision language approach for mobile robot social navigation. [ICRA'25] [Paper]
- Semantic audio-visual navigation. [CVPR'21] [Paper]
- Soundspaces: Audio-visual navigation in 3d environments. [ECCV'20] [Paper]
- VAMOS: A Hierarchical Vision-Language-Action Model for Capability-Modulated and Steerable Navigation. [arXiv'25] [Paper]
- Tell me where to go: A composable framework for context-aware embodied robot navigation. [arXiv'23] [Paper]
- Collaborative instance object navigation: Leveraging uncertainty-awareness to minimize human-agent dialogues. [ICCV'25] [Paper]
- Internscenes: A large-scale simulatable indoor scene dataset with realistic layouts. [arXiv'25] [Paper]
- See, Point, Fly: A Learning-Free VLM Framework for Universal Unmanned Aerial Navigation. [CORL'25] [Paper]
@misc{longxi2026fmnav,
title = {Foundation Models for Embodied Navigation: A Survey},
author = {Gao, Longxi and Xie, Weikai and Qian, Haoze and Yi, Rongjie and Wang, Shihe and Song, Jiaye and Cai, Dongqi and Yuan, Jinliang and Liu, Yunhao and Liu, Xuanzhe and Wang, Shangguang and Xu, Mengwei},
year = {2026},
howpublished = {\url{https://MEmbodied.github.io/embodied-navigation-survey}},
note = {Online; accessed 2026-03-25}
}
If we leave out any important papers, please let us know in the Issues and we will include them in the next version.