Learning to Score Behaviors for Guided Policy Optimization
This work addresses the challenge of guiding policy optimization towards desired behaviors in reinforcement learning, representing an incremental improvement with specific algorithmic contributions.
The paper tackles the problem of comparing reinforcement learning policies by introducing a method using Wasserstein distances in a latent behavioral space, resulting in novel algorithms that outperform existing methods in challenging environments.
We introduce a new approach for comparing reinforcement learning policies, using Wasserstein distances (WDs) in a newly defined latent behavioral space. We show that by utilizing the dual formulation of the WD, we can learn score functions over policy behaviors that can in turn be used to lead policy optimization towards (or away from) (un)desired behaviors. Combined with smoothed WDs, the dual formulation allows us to devise efficient algorithms that take stochastic gradient descent steps through WD regularizers. We incorporate these regularizers into two novel on-policy algorithms, Behavior-Guided Policy Gradient and Behavior-Guided Evolution Strategies, which we demonstrate can outperform existing methods in a variety of challenging environments. We also provide an open source demo.