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

By Mitchell Wortsman, Kiana Ehsani, Mohammad Rastegari, Ali Farhadi, Roozbeh Mottaghi. 2018

264 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Continuous Scene Representations for Embodied AI (2022, background): “Recently, various embodied tasks such as navigation [5,8,37,54,56,63], instruction following [1,47], embodied question answering [15, 23], object manipulation [18,21,22,46], and room rearrangement [4,52] have been of interest.”
AI2-THOR: An Interactive 3D Environment for Visual AI (2017, methodology): “t: http://ai2thor. allenai.org/documentation 7. Example Usage of AI2-THOR AI2-THOR has been successfully tested for different applications. We used AI2-THOR for the problem of targetdriven navigation [26,24,22] using deep reinforcement learning and showed that the trained model can be transferred to”
ObjectNav Revisited: On Evaluation of Embodied Agents Navigating to Objects (2020, methodology): “Both iTHOR and RoboTHOR have been used as platforms for the ObjectNav task [8,20,21].”
CoWs on Pasture: Baselines and Benchmarks for Language-Driven Zero-Shot Object Navigation (2022, methodology): “In this paper, we study Language-driven zero-shot object navigation (L-ZSON)—a more challenging but also more applicable version of object navigation [4,17,63,74,83] and ZSON [37, 44] tasks.”
ProcTHOR: Large-Scale Embodied AI Using Procedural Generation (2022, background): “For instance, in embodied AI, the houses may be used for navigation [58, 89, 118, 132, 117, 128, 74, 130], multi-agent interaction [51, 52, 2], rearrangement and interaction [115, 36, 39, 23, 106], manipulation [33, 86, 32, 122], Sim2Real transfer [27, 54, 66], embodied vision-andlanguage [105, 87, ”
Object Goal Navigation using Goal-Oriented Semantic Exploration (2020, background): “[43] propose a meta-reinforcement learning approach where an agent learns a self-supervised interaction loss that encourages effective navigation to even keep learning in a test environment.”
ActioNet: An Interactive End-To-End Platform For Task-Based Data Collection And Augmentation In 3D Environment (2020, methodology): “Besides reinforcement learning [1], there is increasing interest in datadriven approach [2] [3] [4] for the study of task planning”
SoundSpaces: Audio-Visual Navigation in 3D Environments (2019, methodology): “Our network architecture is inspired by current RL models in the visual navigation literature [59,98,21,47].”
CLIP on Wheels: Zero-Shot Object Navigation as Object Localization and Exploration (2022, background): “We define the task of navigating to unseen goal objects in unseen domains, modifying the notation from [69].”
From Seeing to Moving: A Survey on Learning for Visual Indoor Navigation (VIN) (2020, methodology): “The authors of [Wortsman et al., 2019] adopted a meta-learningbased method to quickly adapt the learned model to the new environments with the proposed self-supervised interaction loss.”
Audio-Visual Embodied Navigation (2019, methodology): “We extend the network architecture used in [61, 102, 23, 49, 24, 98] to also process audio input.”
Learning Object Relation Graph and Tentative Policy for Visual Navigation (2020, methodology): “We use the evaluation protocol in [27].”
Diffusion-based Generation, Optimization, and Planning in 3D Scenes (2023, background): “Among all tasks, visual navigation has been most studied in the vision and robotics community [4, 13, 21, 40, 76, 77, 94, 95].”
Visual Room Rearrangement (2021, background): “Various tasks have been proposed in this domain: navigation towards objects [4, 51, 48, 7] or towards a speciﬁc point [1, 38, 47, 36], scene exploration [9, 8], embodied question answering [18, 13], task completion [55], instruction following [2, 41], object manipulation [16, 52], multi-agent…”
A Cordial Sync: Going Beyond Marginal Policies for Multi-Agent Embodied Tasks (2020, background): “based learning [99,103], meta-learning [98], unimodal baselines [90], 3D point clouds [97], and effective exploration [95,78,16,74] to improve embodied navigational agents.”
RoboTHOR: An Open Simulation-to-Real Embodied AI Platform (2020, methodology): “[69] use meta-learning to learn a self-supervised navigation policy toward a specified object category.”
Exploiting Scene-specific Features for Object Goal Navigation (2020, methodology): “For the Target Driven Navigation some works have recently been presented, such as [30][34][31].”
Pushing it out of the Way: Interactive Visual Navigation (2021, background): “A popular task that has received a considerable amount of attention is visual navigation [3, 5, 8, 29, 39, 48], where the goal is to navigate towards a specific coordinate or object within an unseen environment.”