Visual domain randomization in simulated environments is a widely used method to transfer policies trained in simulation to real robots. However, domain randomization and augmentation hamper the training of a policy. As reinforcement learning struggles with a noisy training signal, this additional nuisance can drastically impede training. For difficult tasks it can even result in complete failure to learn. To overcome this problem we propose to pre-train a perception encoder that already provides an embedding invariant to the randomization. We demonstrate that this yields consistently improved results on a randomized version of DeepMind control suite tasks and a stacking environment on arbitrary backgrounds with zero-shot transfer to a physical robot.
@article{arxiv.2104.14386,
title = {Pre-training of Deep RL Agents for Improved Learning under Domain Randomization},
author = {Artemij Amiranashvili and Max Argus and Lukas Hermann and Wolfram Burgard and Thomas Brox},
journal= {arXiv preprint arXiv:2104.14386},
year = {2021}
}