mlsanityData to 14 Oct 2026

Learning to Learn How to Learn: Self-Adaptive Visual Navigation Using Meta-Learning

arXiv 1812.00971 ↗First posted 3 Dec 2018cs.CV cs.AI cs.LGcited by 93

Not on the map. Why?

Abstract (arXiv)

Learning is an inherently continuous phenomenon. When humans learn a new task there is no explicit distinction between training and inference. As we learn a task, we keep learning about it while performing the task. What we learn and how we learn it varies during different stages of learning. Learning how to learn and adapt is a key property that enables us to generalize effortlessly to new settings. This is in contrast with conventional settings in machine learning where a trained model is frozen during inference. In this paper we study the problem of learning to learn at both training and test time in the context of visual navigation. A fundamental challenge in navigation is generalization to unseen scenes. In this paper we propose a self-adaptive visual navigation method (SAVN) which learns to adapt to new environments without any explicit supervision. Our solution is a meta-reinforcement learning approach where an agent learns a self-supervised interaction loss that encourages effective navigation. Our experiments, performed in the AI2-THOR framework, show major improvements in both success rate and SPL for visual navigation in novel scenes. Our code and data are available at: https://github.com/allenai/savn .

Key facts. Cited by 93 papers in the map (data through 14 Oct 2026). First posted on arXiv 3 Dec 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2020: 1; 2021: 22; 2022: 20; 2023: 22; 2024: 14; 2025: 9; 2026: 5.

93 papers.

Sort:

By area:

  1. 2022ProcTHOR: Large-Scale Embodied AI Using Procedural Generationcited by 380
  2. 2021A Survey of Embodied AI: From Simulators to Research Taskscited by 261
  3. 2023Diffusion-based Generation, Optimization, and Planning in 3D Scenescited by 239
  4. 2023Habitat 3.0: A Co-Habitat for Humans, Avatars and Robotscited by 228
  5. 2022CoWs on Pasture: Baselines and Benchmarks for Language-Driven Zero-Shot Object Navigationcited by 194
  6. 2021Visual Room Rearrangementcited by 102
  7. 2021ManipulaTHOR: A Framework for Visual Object Manipulationcited by 102
  8. 2024GOAT-Bench: A Benchmark for Multi-Modal Lifelong Navigationcited by 94
  9. 2022Zero Experience Required: Plug & Play Modular Transfer Learning for Semantic Visual Navigationcited by 67
  10. 2021RobustNav: Towards Benchmarking Robustness in Embodied Navigationcited by 54
  11. 2021Deep Federated Learning for Autonomous Drivingcited by 54
  12. 2021VTNet: Visual Transformer Network for Object Goal Navigationcited by 51
38 more
  1. 2021Visual Navigation with Spatial Attentioncited by 43
  2. 2021Offline Meta-Reinforcement Learning for Industrial Insertioncited by 39
  3. 2022Continuous Scene Representations for Embodied AIcited by 39
  4. 2021Vision-Language Navigation: A Survey and Taxonomycited by 34
  5. 2021Hierarchical Object-to-Zone Graph for Object Navigationcited by 34
  6. 2022AssistQ: Affordance-centric Question-driven Task Completion for Egocentric Assistantcited by 32
  7. 2022Zero-shot object goal visual navigationcited by 31
  8. 2022TIDEE: Tidying Up Novel Rooms using Visuo-Semantic Commonsense Priorscited by 30
  9. 2022Phone2Proc: Bringing Robust Robots Into Our Chaotic Worldcited by 26
  10. 2022Interactron: Embodied Adaptive Object Detectioncited by 21
  11. 2022Object Memory Transformer for Object Goal Navigationcited by 20
  12. 2022Search for or Navigate to? Dual Adaptive Thinking for Object Navigationcited by 20
  13. 2022Unbiased Directed Object Attention Graph for Object Navigationcited by 18
  14. 2021Pushing it out of the Way: Interactive Visual Navigationcited by 16
  15. 2022Navigating to Objects in Unseen Environments by Distance Predictioncited by 16
  16. 2021On sensitivity of meta-learning to support datacited by 15
  17. 2023Evidential Active Recognition: Intelligent and Prudent Open-World Embodied Perceptioncited by 15
  18. 2024TopV-Nav: Unlocking the Top-View Spatial Reasoning Potential of MLLM for Zero-shot Object Navigationcited by 15
  19. 2020Towards Target-Driven Visual Navigation in Indoor Scenes via Generative Imitation Learningcited by 14
  20. 2021Towards Optimal Correlational Object Searchcited by 14
  21. 2023Learning to View: Decision Transformers for Active Object Detectioncited by 11
  22. 2021Multi-Agent Embodied Visual Semantic Navigation with Scene Prior Knowledgecited by 10
  23. 2021Agent-Centric Relation Graph for Object Visual Navigationcited by 10
  24. 2025Nav-R2 Dual-Relation Reasoning for Generalizable Open-Vocabulary Object-Goal Navigationcited by 10
  25. 2024Aligning Knowledge Graph with Visual Perception for Object-goal Navigationcited by 9
  26. 2025PanoNav: Mapless Zero-Shot Object Navigation with Panoramic Scene Parsing and Dynamic Memorycited by 9
  27. 2022What do navigation agents learn about their environment?cited by 8
  28. 2023Multiple Thinking Achieving Meta-Ability Decoupling for Object Navigationcited by 8
  29. 2023A Partially Supervised Reinforcement Learning Framework for Visual Active Searchcited by 8
  30. 2024HELPER-X: A Unified Instructable Embodied Agent to Tackle Four Interactive Vision-Language Domains with Memory-Augmented Language Modelscited by 8
  31. 2024Advancing DRL Agents in Commercial Fighting Games: Training, Integration, and Agent-Human Alignmentcited by 8
  32. 2022Embodied Referring Expression for Manipulation Question Answering in Interactive Environmentcited by 7
  33. 2023Multi-Level Compositional Reasoning for Interactive Instruction Followingcited by 7
  34. 2024TDANet: Target-Directed Attention Network For Object-Goal Visual Navigation With Zero-Shot Abilitycited by 7
  35. 2025GS-LTS: 3D Gaussian Splatting-Based Adaptive Modeling for Long-Term Service Robotscited by 7
  36. 2022Object Manipulation via Visual Target Localizationcited by 6
  37. 2023Heterogeneous Embodied Multi-Agent Collaborationcited by 6
  38. 2023Active Open-Vocabulary Recognition: Let Intelligent Moving Mitigate CLIP Limitationscited by 6

50 most cited of 93 All 93 (JSON)