This paper describes the adaption and application of an algorithm called Feudal Reinforcement Learning to a complex gridworld navigation problem. The algorithm proved to be not easily adaptable and the results were unsatisfactory.
Peter Dayan, Geoffrey E. Hinton