Multi-level dynamics of a simulated agent system

Simulated multi-agent football provides a good test-bed to study the properties and dynamics of multi-level systems. A single player can be considered as the basic building block of the system, then groups of players form structures which can be considered to exit at a higher level than the players themselves. Structures have properties that single players do not, such as being strategically good. This paper presents a network based approach to represent and study the properties and dynamics of multi-level systems. In this paper the elements of a system (i.e. football agents, ball and goals) are represented by network nodes and their interactions as edges. A fundamental idea in this paper is that interconnected nodes are considered as new constructs that exist one level above the elements and interactions that de¯ne them. This assembly of networks into higher-level constructs leads into a multi-level representation of the simulated football system.