Embodied Visual Navigation with Automatic Curriculum Learning in Real Environments
This addresses the challenge of efficient and transferable navigation for robots in real-world settings, though it appears incremental as it builds on existing curriculum learning methods.
The paper tackles the problem of training embodied agents for visual navigation in unknown cluttered indoor environments using only RGB images, and results show that agents trained with NavACL significantly outperform state-of-the-art agents trained with uniform sampling.
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.