Planning in Hierarchical Reinforcement Learning: Guarantees for Using Local Policies
This work addresses the challenge of efficient policy composition in hierarchical RL for settings with collectible rewards, offering incremental improvements in worst-case performance guarantees.
The paper tackles the problem of assembling a global policy from local reward-maximizing policies in hierarchical reinforcement learning with sum-of-components rewards, providing theoretical guarantees for deterministic MDPs and proposing three local stochastic policies that outperform deterministic ones in worst-case scenarios.
We consider a settings of hierarchical reinforcement learning, in which the reward is a sum of components. For each component we are given a policy that maximizes it and our goal is to assemble a policy from the individual policies that maximizes the sum of the components. We provide theoretical guarantees for assembling such policies in deterministic MDPs with collectible rewards. Our approach builds on formulating this problem as a traveling salesman problem with discounted reward. We focus on local solutions, i.e., policies that only use information from the current state; thus, they are easy to implement and do not require substantial computational resources. We propose three local stochastic policies and prove that they guarantee better performance than any deterministic local policy in the worst case; experimental results suggest that they also perform better on average.