Detection of emergent phenomena in multi-agent systems

Biological systems are composed of a large number of molecules. These systems are complex because numerous and divesified interactions exist between the molecules making their understanding difficult. Due to these interactions, complex systems can show some phenomena considered as emergents because unpredictable from the simple description of molecules' behaviour. We particularly study self-organisation phenomena corresponding to a structural organisation without central control. Many tools are available to model these biological complex systems. We use multi-agent systems which model each element of the system as an agent. We study their collective behaviour inside their environment. Multi-agent simulations exhibit phenomena which appear in the real systems they model. The characterisation of these phenomena is usually given by the users observing such systems, which introduces an important subjective bias. The objective of our work is to automatically detect some emergent phenomena. We model multi-agent simulation as a graph where vertices represent agents and edges their interactions. This graph is modified at each time step and we study its properties