State Space Decomposition and Subgoal Creation for Transfer in Deep Reinforcement Learning
This addresses the issue of domain-specific policies in RL for researchers and practitioners, though it appears incremental as it builds on existing attention-based methods.
The paper tackles the problem of poor generalization in deep reinforcement learning by developing a framework that uses a recurrent attention mechanism to decompose state spaces and create subgoals, enabling transfer from simpler to more complex domains, with experiments showing the meta-controller successfully learns to generate these subgoals.
Typical reinforcement learning (RL) agents learn to complete tasks specified by reward functions tailored to their domain. As such, the policies they learn do not generalize even to similar domains. To address this issue, we develop a framework through which a deep RL agent learns to generalize policies from smaller, simpler domains to more complex ones using a recurrent attention mechanism. The task is presented to the agent as an image and an instruction specifying the goal. This meta-controller guides the agent towards its goal by designing a sequence of smaller subtasks on the part of the state space within the attention, effectively decomposing it. As a baseline, we consider a setup without attention as well. Our experiments show that the meta-controller learns to create subgoals within the attention.