Decentralized planning

Central coordination cannot always provide a satisfying solution as we explained before. An alternative might be decentralized planning.

With decentralized control we mean the coordination of activities among multiple self-interested players that make decisions independently and where the system performance is a result of the actions and decisions of individuals. Centralized control, on the other hand, is the coordination of activities by a single party that decides on and coordinates the actions of individuals, to maximize the system performance. Decentralized control is in the literature also referred to as distributed rational decision making or just Multi-Agent systems.

The idea of decentralized coordination can also be found in nature. If we observe ants carefully, we do not see any form of central coordination. There is no planning department or a chief-ant determining where food or nest material is collected and who is put into action. Each ant works autonomously, has bounded rationality, limited physical abilities, limited knowledge of its surrounding, and limited possibilities to communicate with other ants. Yet, they are able to organize themselves effectively.

Something similar we aim to realize by using Multi-Agent systems. We try to solve a problem through the interaction of self-interested and autonomous players by providing interaction protocols that help the players to organize themselves effectively. Next>>