Why Generalization in RL is Difficult: Epistemic POMDPs and Implicit Partial Observability
This addresses a central problem for deploying RL systems in real-world settings, offering a novel perspective but with incremental algorithmic improvements.
The paper tackles the challenge of generalization in reinforcement learning (RL) by showing that generalization to unseen conditions induces implicit partial observability, turning fully-observed MDPs into POMDPs, and proposes an ensemble-based technique that achieves significant gains on the Procgen benchmark.
Generalization is a central challenge for the deployment of reinforcement learning (RL) systems in the real world. In this paper, we show that the sequential structure of the RL problem necessitates new approaches to generalization beyond the well-studied techniques used in supervised learning. While supervised learning methods can generalize effectively without explicitly accounting for epistemic uncertainty, we show that, perhaps surprisingly, this is not the case in RL. We show that generalization to unseen test conditions from a limited number of training conditions induces implicit partial observability, effectively turning even fully-observed MDPs into POMDPs. Informed by this observation, we recast the problem of generalization in RL as solving the induced partially observed Markov decision process, which we call the epistemic POMDP. We demonstrate the failure modes of algorithms that do not appropriately handle this partial observability, and suggest a simple ensemble-based technique for approximately solving the partially observed problem. Empirically, we demonstrate that our simple algorithm derived from the epistemic POMDP achieves significant gains in generalization over current methods on the Procgen benchmark suite.