Towards Run-time Efficient Hierarchical Reinforcement Learning

This paper investigates a novel method combining Scalable Evolution Strategies (S-ES) and Hierarchical Reinforcement Learning (HRL). S-ES, named for its excellent scalability, was popularised with demonstrated performance comparable to state-of-the-art policy gradient methods. However, S-ES has not been tested in conjunction with HRL methods, which empower temporal abstraction thus allowing agents to tackle more challenging problems. We introduce a novel method merging S-ES and HRL, which creates a highly scalable and efficient (compute time) algorithm. We demonstrate that the proposed method benefits from S-ES's scalability and indifference to delayed rewards. This results in our main contribution: significantly higher learning speed and competitive performance compared to gradient-based HRL methods, across a range of tasks.