Processing math: 100%
Bi-Directional Domain Adaptation for Sim2Real Transfer of Embodied Navigation Agents | IEEE Journals & Magazine | IEEE Xplore

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

Publisher: IEEE

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...View more

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 letter, 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 5 k real-world (state, action, next-state) samples matches the performance of a policy fine-tuned with \sim 600 k samples, resulting in a speed-up of \sim 120\times .
Published in: IEEE Robotics and Automation Letters ( Volume: 6, Issue: 2, April 2021)
Page(s): 2634 - 2641
Date of Publication: 25 February 2021

ISSN Information:

Publisher: IEEE

Funding Agency:


References

References is not available for this document.