Hierarchical Behavior Cloning

[Dhanushvarma]

Problem

Implement an hierarchical Behavior Cloning Algorithms

Solution

Build on top of the existing BC code, modified to have two main components: 1)High level planner for intermediate goals 2) low level goal (set by the high level planner). The actual neural network architectures used for both these is up to choice.

I wanted to implement this using the "imitation" library , do the author's suggest that the library provides support for this kind of an algorithm.