Pre-training of Deep RL Agents for Improved Learning under Domain Randomization
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 n...
Gespeichert in:
Hauptverfasser: | , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | 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. |
---|---|
DOI: | 10.48550/arxiv.2104.14386 |