Learning Exploration Policies for Navigation

Abstract

Numerous past works have tackled the problem of task-driven navigation. But, how to effectively explore a new environment to enable a variety of down-stream tasks has received much less attention. In this work, we study how agents can autonomously explore realistic and complex 3D environments without the context of task-rewards. We propose a learning-based approach and investigate different policy architectures, reward functions, and training paradigms. We find that the use of policies with spatial memory that are bootstrapped with imitation learning and finally finetuned with coverage rewards derived purely from on-board sensors can be effective at exploring novel environments. We show that our learned exploration policies can explore better than classical approaches based on geometry alone and generic learning-based exploration techniques. Finally, we also show how such task-agnostic exploration can be used for down-stream tasks. Code and Videos are available at: https://sites.google.com/view/exploration-for-nav.

References in corpus (7)

• RL${}^2$: Fast Reinforcement Learning via Slow Reinforcement Learning
• On Evaluation of Embodied Navigation Agents
• Learning to Navigate in Complex Environments
• Incentivizing Exploration In Reinforcement Learning With Deep Predictive Models
• Building Generalizable Agents with a Realistic and Rich 3D Environment
• Neural SLAM: Learning to Explore with External Memory
• EX2: Exploration with Exemplar Models for Deep Reinforcement Learning

Cited by in corpus (15)

• Cross-view Semantic Segmentation for Sensing Surroundings
• Spatial Action Maps for Mobile Manipulation
• Mid-Level Visual Representations Improve Generalization and Sample Efficiency for Learning Visuomotor Policies
• Emergence of Exploratory Look-Around Behaviors through Active Observation Completion
• A Survey of Embodied AI: From Simulators to Research Tasks
• Learning Object Relation Graph and Tentative Policy for Visual Navigation
• Integrating Egocentric Localization for More Realistic Point-Goal Navigation Agents
• Navigate-and-Seek: a Robotics Framework for People Localization in Agricultural Environments
• Topological Planning with Transformers for Vision-and-Language Navigation
• Simultaneous Mapping and Target Driven Navigation
• MVP: Unified Motion and Visual Self-Supervised Learning for Large-Scale Robotic Navigation
• Learning to Move with Affordance Maps
• Unsupervised Domain Adaptation for Visual Navigation
• Learning to Map for Active Semantic Goal Navigation
• Learning to Explore by Reinforcement over High-Level Options