Reward Scale Robustness for Proximal Policy Optimization via DreamerV3 Tricks
This work addresses the robustness of reinforcement learning methods to reward scales for researchers, showing incremental results by testing existing tricks on a different algorithm.
The study applied DreamerV3's tricks to PPO to test their generality, finding they do not generally improve PPO, with experiments totaling over 10,000 A100 hours on benchmarks like Arcade Learning Environment and DeepMind Control Suite.
Most reinforcement learning methods rely heavily on dense, well-normalized environment rewards. DreamerV3 recently introduced a model-based method with a number of tricks that mitigate these limitations, achieving state-of-the-art on a wide range of benchmarks with a single set of hyperparameters. This result sparked discussion about the generality of the tricks, since they appear to be applicable to other reinforcement learning algorithms. Our work applies DreamerV3's tricks to PPO and is the first such empirical study outside of the original work. Surprisingly, we find that the tricks presented do not transfer as general improvements to PPO. We use a high quality PPO reference implementation and present extensive ablation studies totaling over 10,000 A100 hours on the Arcade Learning Environment and the DeepMind Control Suite. Though our experiments demonstrate that these tricks do not generally outperform PPO, we identify cases where they succeed and offer insight into the relationship between the implementation tricks. In particular, PPO with these tricks performs comparably to PPO on Atari games with reward clipping and significantly outperforms PPO without reward clipping.