Replication of the first experiment of Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation (Kulkarni et al., 2016) (view here).
Download the report here.
View the presentation here.
This work was done as a class project for MIT 6.882: Embodied Intelligence. Credit is due to Professor Tomas Lozano-Perez for providing valuable feedback on my approach. Credit is also due to a previous replication attempt of the hierarchical-DQN paper, which did not successfully replicate the results but inspired aspects of this implementation: https://github.com/EthanMacdonald/h-DQN