Meta Learning Shared Hierarchies
This addresses sample efficiency for robotics tasks where learning from scratch is impractical, though it builds incrementally on hierarchical RL and meta-learning concepts.
The paper tackles the problem of sample inefficiency in reinforcement learning for unseen tasks by developing a meta-learning approach that learns hierarchically structured policies with shared primitives. The method successfully discovers meaningful motor primitives for four-legged robots in maze distributions and enables transfer to long-timescale sparse-reward obstacle courses and 3D humanoid walking/crawling.
We develop a metalearning approach for learning hierarchically structured policies, improving sample efficiency on unseen tasks through the use of shared primitives---policies that are executed for large numbers of timesteps. Specifically, a set of primitives are shared within a distribution of tasks, and are switched between by task-specific policies. We provide a concrete metric for measuring the strength of such hierarchies, leading to an optimization problem for quickly reaching high reward on unseen tasks. We then present an algorithm to solve this problem end-to-end through the use of any off-the-shelf reinforcement learning method, by repeatedly sampling new tasks and resetting task-specific policies. We successfully discover meaningful motor primitives for the directional movement of four-legged robots, solely by interacting with distributions of mazes. We also demonstrate the transferability of primitives to solve long-timescale sparse-reward obstacle courses, and we enable 3D humanoid robots to robustly walk and crawl with the same policy.