The research proposes Adaptive Bayesian Domain Randomization via Strategic Fine-tuning (BayRnTune), a method to significantly speed up the learning processes in domain randomization by fine-tuning from previously learned policy. The study investigates four different fine-tuning strategies and compares them against baseline algorithms in five simulated environments. The analysis shows that this method gives better rewards in the same amount of timesteps compared to vanilla domain randomization or Bayesian Domain Randomization.
Publication date: 18 Oct 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2310.10606
