Hierarchical Control in a Multiagent System
Ariuna Damba, Shigeyoshi Watanabe · 2007
The hierarchical decomposition approach as a pervasive method to deal with undesirable computation allows for reduced computation, and provides solutions for larger and more complicated problems. However, there is a necessity for research efforts towards the arrangement of local operational performance criteria to the overall performance of a whole system. This paper uses the method described in our previous works on decision making simulations based on computed policy functions in a multiagent system. We propose that a multiagent system design for efficient control with hierarchical representation, is closely related to agent modeling features. Considering the individual and social criterias while retaining the benefits of hierarchical coordination among agents, is the challenging issue of this paper. An experimental study with a game model shows that systems more sensitive in terms of computational resources can take advantages of hierarchical control design, where individual and social criterias play significant role in the system design mechanism.