Deep Learning for Embodied Vision Navigation: A Survey
arXiv:2108.04097
Abstract
"Embodied visual navigation" problem requires an agent to navigate in a 3D environment mainly rely on its first-person observation. This problem has attracted rising attention in recent years due to its wide application in autonomous driving, vacuum cleaner, and rescue robot. A navigation agent is supposed to have various intelligent skills, such as visual perceiving, mapping, planning, exploring and reasoning, etc. Building such an agent that observes, thinks, and acts is a key to real intelligence. The remarkable learning ability of deep learning methods empowered the agents to accomplish embodied visual navigation tasks. Despite this, embodied visual navigation is still in its infancy since a lot of advanced skills are required, including perceiving partially observed visual input, exploring unseen areas, memorizing and modeling seen scenarios, understanding cross-modal instructions, and adapting to a new environment, etc. Recently, embodied visual navigation has attracted rising attention of the community, and numerous works has been proposed to learn these skills. This paper attempts to establish an outline of the current works in the field of embodied visual navigation by providing a comprehensive literature survey. We summarize the benchmarks and metrics, review different methods, analysis the challenges, and highlight the state-of-the-art methods. Finally, we discuss unresolved challenges in the field of embodied visual navigation and give promising directions in pursuing future research.
arXiv comment20 pages
References in corpus (30)
- Sequence to Sequence Learning with Neural Networks
- BioBERT: a pre-trained biomedical language representation model for biomedical text mining
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
- Joint 2D-3D-Semantic Data for Indoor Scene Understanding
- On Evaluation of Embodied Navigation Agents
- Learning to Navigate in Complex Environments
- Adaptive Computation Time for Recurrent Neural Networks
- AI2-THOR: An Interactive 3D Environment for Visual AI
- Reinforcement Learning with Unsupervised Auxiliary Tasks
- Lifelong Federated Reinforcement Learning: A Learning Architecture for Navigation in Cloud Robotic Systems
- Learning to Explore using Active Neural SLAM
- Comparison of Various SLAM Systems for Mobile Robot in an Indoor Environment
- MINOS: Multimodal Indoor Simulator for Navigation in Complex Environments
- Neural SLAM: Learning to Explore with External Memory
- Sim-to-Real Robot Learning from Pixels with Progressive Nets
- Video Question Answering via Attribute-Augmented Attention Network Learning
- Policy Distillation
- Towards Cognitive Exploration through Deep Reinforcement Learning for Mobile Robots
- Learning to Act by Predicting the Future
- One-Shot Reinforcement Learning for Robot Navigation with Interactive Replay
- MultiON: Benchmarking Semantic Map Memory using Multi-Object Navigation
- iGibson 1.0: a Simulation Environment for Interactive Tasks in Large Realistic Scenes
- Embodied Visual Navigation with Automatic Curriculum Learning in Real Environments
- Blindfold Baselines for Embodied QA
- General Evaluation for Instruction Conditioned Navigation using Dynamic Time Warping
- MQA: Answering the Question via Robotic Manipulation
- Improving Target-driven Visual Navigation with Attention on 3D Spatial Relationships
- Integrating Egocentric Localization for More Realistic Point-Goal Navigation Agents
- Multi-View Learning for Vision-and-Language Navigation
- Learning to Move with Affordance Maps