Embodied Visual Navigation with Automatic Curriculum Learning in Real Environments

TL;DR

This paper introduces NavACL, an automatic curriculum learning method for embodied visual navigation that improves training efficiency and enables real-world navigation of cluttered environments using only RGB images.

Contribution

NavACL provides a simple, geometric feature-based curriculum learning approach that enhances deep reinforcement learning for navigation tasks, outperforming uniform sampling methods.

Findings

NavACL-trained agents outperform state-of-the-art uniform sampling agents.

Agents can navigate unknown cluttered environments using only RGB images.

Policies transfer seamlessly to real-world robots without retraining.

Abstract

We present NavACL, a method of automatic curriculum learning tailored to the navigation task. NavACL is simple to train and efficiently selects relevant tasks using geometric features. In our experiments, deep reinforcement learning agents trained using NavACL significantly outperform state-of-the-art agents trained with uniform sampling -- the current standard. Furthermore, our agents can navigate through unknown cluttered indoor environments to semantically-specified targets using only RGB images. Obstacle-avoiding policies and frozen feature networks support transfer to unseen real-world environments, without any modification or retraining requirements. We evaluate our policies in simulation, and in the real world on a ground robot and a quadrotor drone. Videos of real-world results are available in the supplementary material.