Multi-Agent Systems and Heterogeneous Scales Interactions. Application to Pharmacokinetics of Vitamin K Antagonists

The study of complex systems consists in considering entities submitted to interactions which define the dynamics of the system. Virtual reality opens the way to interactive simulation of complex systems, so called the in virtuo experimentation. For that purpose we use multi-interactions systems, based on the reification of interactions and multi-agent systems, in a phenomenological approach. Interaction agents represent the modeler understanding of the relations between the constituents of the system. Such descriptive models lead us to define parameters a priori. Moreover these parameters can be fluctuant, or even unknown, during a simulation in relation to the system dynamics or user interventions. To respond to this problem, we expose in this paper a redundant multiscale architecture which rests upon the fact that we can establish models of a same phenomenon at heterogeneous time and space scales. Heterogenous Multiscale Methods provide a general framework to mix levels of description of a system. Our intention is to implement this framework in multi-interactions systems by means of a Scale-Interaction agent. Then we illustrate our architecture through a pharmacokinetics application. Indeed biochemical kinetics abounds of parametric phenomena. Finally we discuss about some questions raised by this methodology, such as synchronicity, organization detection and genericity