Safe Reinforcement Learning From Pixels Using a Stochastic Latent Representation
This work addresses the problem of ensuring safety in reinforcement learning for agents operating in high-dimensional, partially observable environments, representing an incremental improvement over existing methods.
The paper tackles safe reinforcement learning from pixel observations by addressing the trade-off between reward optimization and safety constraints, partial observability, and high-dimensional observations, demonstrating competitive performance in benchmark environments with respect to computational requirements, reward return, and safety constraint satisfaction.
We address the problem of safe reinforcement learning from pixel observations. Inherent challenges in such settings are (1) a trade-off between reward optimization and adhering to safety constraints, (2) partial observability, and (3) high-dimensional observations. We formalize the problem in a constrained, partially observable Markov decision process framework, where an agent obtains distinct reward and safety signals. To address the curse of dimensionality, we employ a novel safety critic using the stochastic latent actor-critic (SLAC) approach. The latent variable model predicts rewards and safety violations, and we use the safety critic to train safe policies. Using well-known benchmark environments, we demonstrate competitive performance over existing approaches with respects to computational requirements, final reward return, and satisfying the safety constraints.