Bi-directional Domain Adaptation for Sim2Real Transfer of Embodied Navigation Agents

TL;DR

This paper introduces Bi-directional Domain Adaptation (BDA), a method to bridge the sim-vs-real gap in embodied navigation by adapting both visual and dynamics domains, significantly reducing real-world data requirements.

Contribution

The paper presents a novel BDA approach that simultaneously addresses visual and dynamics domain gaps, enabling efficient sim2real transfer for navigation agents.

Findings

BDA with 5k real samples matches 600k fine-tuned samples.

BDA achieves a 120x speed-up in training efficiency.

Effective in PointGoal Navigation tasks.

Abstract

Deep reinforcement learning models are notoriously data hungry, yet real-world data is expensive and time consuming to obtain. The solution that many have turned to is to use simulation for training before deploying the robot in a real environment. Simulation offers the ability to train large numbers of robots in parallel, and offers an abundance of data. However, no simulation is perfect, and robots trained solely in simulation fail to generalize to the real-world, resulting in a "sim-vs-real gap". How can we overcome the trade-off between the abundance of less accurate, artificial data from simulators and the scarcity of reliable, real-world data? In this paper, we propose Bi-directional Domain Adaptation (BDA), a novel approach to bridge the sim-vs-real gap in both directions -- real2sim to bridge the visual domain gap, and sim2real to bridge the dynamics domain gap. We demonstrate the benefits of BDA on the task of PointGoal Navigation. BDA with only 5k real-world (state, action, next-state) samples matches the performance of a policy fine-tuned with ~600k samples, resulting in a speed-up of ~120x.