Feature-Attending Recurrent Modules for Generalization in Reinforcement Learning
This addresses the challenge of generalization in reinforcement learning for object-centric tasks, representing an incremental improvement over prior methods.
The paper tackles the problem of enabling reinforcement learning agents to generalize across object-centric tasks by introducing Feature-Attending Recurrent Modules (FARM), an architecture that learns state representations with inductive biases for spatial and temporal regularities, resulting in improved generalization compared to competing architectures in 2D and 3D environments.
Many important tasks are defined in terms of object. To generalize across these tasks, a reinforcement learning (RL) agent needs to exploit the structure that the objects induce. Prior work has either hard-coded object-centric features, used complex object-centric generative models, or updated state using local spatial features. However, these approaches have had limited success in enabling general RL agents. Motivated by this, we introduce "Feature-Attending Recurrent Modules" (FARM), an architecture for learning state representations that relies on simple, broadly applicable inductive biases for capturing spatial and temporal regularities. FARM learns a state representation that is distributed across multiple modules that each attend to spatiotemporal features with an expressive feature attention mechanism. We show that this improves an RL agent's ability to generalize across object-centric tasks. We study task suites in both 2D and 3D environments and find that FARM better generalizes compared to competing architectures that leverage attention or multiple modules.