A Nonlinear Multi-agent System designed for Swarm Intelligence: the Logistic MAS

Rodolphe Charrier, Christine Bourjot, Francois Charpillet
2007 First International Conference on Self-Adaptive and Self-Organizing Systems (SASO 2007)  
Ant algorithms and flocking algorithms are the two main programming paradigms in swarm intelligence. They are built on stochastic models, widely used in optimization problems. However, though this modeling leads to highperformance algorithms, some mechanisms, like the symmetry break in ant decision, are still not well understood at the local ant level. Moreover, there is currently no modeling approach which joins the two paradigms. This paper proposes an entirely novel approach to the
more » ... ch to the mathematical foundations of swarm algorithms: contrary to the current stochastic approaches, we show that an alternative deterministic model exists, which has its origin in deterministic chaos theory. We establish a reactive multi-agent system, based on logistic nonlinear decision maps, and designed according to the influence-reaction scheme. The rewriting of the decision functions leads to a new way of understanding the swarm phenomena in terms of state synchronization, and enables the analysis of their convergence behavior through bifurcation diagrams. We apply our approach on two concrete examples of each algorithm class, in order to demonstrate its general applicability.
doi:10.1109/saso.2007.1 dblp:conf/saso/CharrierBC07 fatcat:qgaewnsdkva6hl62oy2xr4xm4m