Memory-Augmented Reinforcement Learning for Image-Goal Navigation
Authors:
- 23 October 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 3316-3323
Abstract
In this work, we present a memory-augmented approach for image-goal navigation. Earlier attempts, including RL-based and SLAM-based approaches have either shown poor generalization performance, or are heavily-reliant on pose/depth sensors. Our method is based on an attention-based end-to-end model that leverages an episodic memory to learn to navigate. First, we train a state-embedding network in a self-supervised fashion, and then use it to embed previously-visited states into the agent's memory. Our navigation policy takes advantage of this information through an attention mechanism. We validate our approach with extensive evaluations, and show that our model establishes a new state of the art on the challenging Gibson dataset. Furthermore, we achieve this impressive performance from RGB input alone, without access to additional information such as position or depth, in stark contrast to related work.Keywords
This publication has 20 references indexed in Scilit:
- Habitat: A Platform for Embodied AI Research2019
- A Behavioral Approach to Visual Navigation with Graph Localization Networks2019
- Scene Memory Transformer for Embodied Agents in Long-Horizon Tasks2019
- Gibson Env: Real-World Perception for Embodied Agents2018
- Learning to Look Around: Intelligently Exploring Unseen Environments for Unknown Tasks2018
- Cognitive Mapping and Planning for Visual Navigation2017
- Target-driven visual navigation in indoor scenes using deep reinforcement learning2017
- ORB-SLAM: A Versatile and Accurate Monocular SLAM SystemIEEE Transactions on Robotics, 2015
- Probabilistic roboticsCommunications of the ACM, 2002
- Topological simultaneous localization and mapping (SLAM): toward exact localization without explicit localizationIEEE Transactions on Robotics and Automation, 2001