Here is some neuroscience work that relates to hierarchical reinforcement learning. Competitive Queing (CQ) is a neuroscience-based concept that I think has similarity to HRL. Learning and utilizing ordinal sets of actions is critical for AGI and many narrow applications of machine learning.
Here is an interesting 2006 presentation on CQ by Dr. Tom Hartley, senior lecturer at York University.
Here’s a paper on the processes.
The paper by Bullock outlines the following steps in the process.
- Sequence Preparation
- Velocity Scaling
- State Sensitive Timing