Autonomy based modeling for the simulation of ocean remote sensing

This article presents the simulation structure for the MODENA project [1] and illustrates what should become an original dynamical model for the reflectivity map of the sea surface. From our point of view, MODENA challenges to find accurate models for simulating dynamically interaction between sea state, sonar, radar and ship, while keeping physical coherency during the whole observation scenario (several tens of minutes), including human participation in this virtual reality system. To deal with such a complexity, as it is impossible and inappropriate to compute dynamically hydrodynamic, electromagnetic and acoustic equations on an accurate grid of points or mesh, MODENA simulation structure aims to use an autonomy based approach: the model of each phenomenon (wave, breaking, group, ship, wind, stream, transmitter, receiver...) involved in the simulation is seen as an autonomous entity, including an autonomy of time, space and scale. These models are combined and animated using enaction-based multi-agents simulation. In such simulation, computer activity is optimised at places where and when interaction occurs within revelant scale, according to choosen models. Furthermore, to add new phenomena or to modify existing models do not make previous work obsolete, as one has only to characterise interactions of new phenomena with previously defined models. This modeling approach facilitates interaction between research teams involved in the MODENA project, as the whole simulation results from the set of autonomized models developed by each team in parallel. Such a virtual laboratory should help for example to better distinguish the signal of breaking from the signal of small boats, in the dynamical signature of sea-states.