Multi-agent systems with many interacting agents require a form of structure or organisation imposed on the population.
Sometimes, this structure is permanent, such as a hierarchy that determines the delegation of control and
propagation of information, or transient, such as a coalition in which agents interact until a certain common goal is
reached. The system designer has to decide which organisations are suitable for the system to reach the stated goals
and implement mechanisms that allow the formation of these organisational structures at runtime. If this process is
driven from within the system, Self-Organisation is present.
This practice includes tasks, work products, and guidance that support the decision for a suitable system structure and
the selection of a suitable self-organisation mechanism. If the system under development requires self-organisation,
e.g., to be robust against agent failures or to adapt to a changing environment, these issues will have to be
considered timely and thoroughly as the organisational structure and the algorithm to create it can have tremendous
impact on the performance of the system. Introducing these concepts also influences the way the system is tested and
deployed and has consequences for the operation of the deployed system. These issues are regarded in more detail in Impact of Self-Organisation on Testing, Deployment, and Operation.
|