Deep Variational Reinforcement Learning for POMDPs
This addresses the challenge of sequential decision-making in noisy, incomplete real-world settings for AI agents, representing an incremental improvement over existing methods.
The paper tackles the problem of reinforcement learning in partially observable environments with unknown models by proposing deep variational reinforcement learning (DVRL), which learns a generative environment model and performs inference to aggregate information, and shows it outperforms previous recurrent neural network approaches on Mountain Hike and flickering Atari benchmarks.
Many real-world sequential decision making problems are partially observable by nature, and the environment model is typically unknown. Consequently, there is great need for reinforcement learning methods that can tackle such problems given only a stream of incomplete and noisy observations. In this paper, we propose deep variational reinforcement learning (DVRL), which introduces an inductive bias that allows an agent to learn a generative model of the environment and perform inference in that model to effectively aggregate the available information. We develop an n-step approximation to the evidence lower bound (ELBO), allowing the model to be trained jointly with the policy. This ensures that the latent state representation is suitable for the control task. In experiments on Mountain Hike and flickering Atari we show that our method outperforms previous approaches relying on recurrent neural networks to encode the past.