Learning Markov State Abstractions for Deep Reinforcement Learning
arXiv:2106.04379
Abstract
A fundamental assumption of reinforcement learning in Markov decision processes (MDPs) is that the relevant decision process is, in fact, Markov. However, when MDPs have rich observations, agents typically learn by way of an abstract state representation, and such representations are not guaranteed to preserve the Markov property. We introduce a novel set of conditions and prove that they are sufficient for learning a Markov abstract state representation. We then describe a practical training procedure that combines inverse model estimation and temporal contrastive learning to learn an abstraction that approximately satisfies these conditions. Our novel training objective is compatible with both online and offline training: it does not require a reward signal, but agents can capitalize on reward information when available. We empirically evaluate our approach on a visual gridworld domain and a set of continuous control benchmarks. Our approach learns representations that capture the underlying structure of the domain and lead to improved sample efficiency over state-of-the-art deep reinforcement learning with visual features -- often matching or exceeding the performance achieved with hand-designed compact state information.
arXiv commentFixed typo (see Errata). Code available at https://github.com/camall3n/markov-state-abstractions
References in corpus (16)
- Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
- Exploration by Random Network Distillation
- dm_control: Software and Tasks for Continuous Control
- Transfer from Simulation to Real World through Learning Deep Inverse Dynamics Model
- Near Optimal Behavior via Approximate State Abstraction
- When to use parametric models in reinforcement learning?
- Loss is its own Reward: Self-Supervision for Reinforcement Learning
- Learning Plannable Representations with Causal InfoGAN
- Decoupling Representation Learning from Reinforcement Learning
- Contingency-Aware Exploration in Reinforcement Learning
- Provably efficient RL with Rich Observations via Latent State Decoding
- Efficient Model-Based Deep Reinforcement Learning with Variational State Tabulation
- Reinforcement Learning with Prototypical Representations
- Decoupling Dynamics and Reward for Transfer Learning
- Successor Features Combine Elements of Model-Free and Model-based Reinforcement Learning
- Deep Radial-Basis Value Functions for Continuous Control