Shaping Belief States with Generative Environment Models for RL
This addresses the challenge of efficient belief-state learning for agents in visually rich environments, though it is incremental as it builds on existing generative model and RL methods.
The paper tackles the problem of training expressive generative models for belief-state formation in complex environments, showing that these models improve data-efficiency in reinforcement learning tasks compared to model-free baselines.
When agents interact with a complex environment, they must form and maintain beliefs about the relevant aspects of that environment. We propose a way to efficiently train expressive generative models in complex environments. We show that a predictive algorithm with an expressive generative model can form stable belief-states in visually rich and dynamic 3D environments. More precisely, we show that the learned representation captures the layout of the environment as well as the position and orientation of the agent. Our experiments show that the model substantially improves data-efficiency on a number of reinforcement learning (RL) tasks compared with strong model-free baseline agents. We find that predicting multiple steps into the future (overshooting), in combination with an expressive generative model, is critical for stable representations to emerge. In practice, using expressive generative models in RL is computationally expensive and we propose a scheme to reduce this computational burden, allowing us to build agents that are competitive with model-free baselines.