Recent advances in hierarchical reinforcement learning (Q5917501)

This paper reviews several closely related approaches to temporal abstraction and hierarchical control that have been developed by machine learning researchers: the options formalism of Sutton, Precup, and Singh, the hierarchies of abstract machines (HAMs) approach of Parr and Russell, and Dietterich's MAXQ framework. The authors also discuss extensions of these ideas addressing on concurrent activities, multiagent coordination, and hierarchical memory for partial observability.