Towards virtual experiment laboratories: How multi-agent simulations can cope with multiple scales of analysis and viewpoints

Abstract

When studying complex phenomena, we face huge diculties to conceive, understand, not to say handle the synthesis process which, from many interacting events, produces an emerging, recognizable, per-sistent and structurally stable, macroscopic event. Such a topical issue calls for specic tools, among which the development of multi-agent sim-ulations has proved a promising approach. However, current multi-agent simulations provide no means of manipulating as a whole dynamically created groups of entities which emerge at dierent granularity levels. To our mind, giving full a sense to multi-agent simulations would con-sist though in making use of such potential groups, by granting them an existence of their own and specic behaviours, thus providing means of apprehending micro-macro links within simulations. We present here a conceptual reflexion on such an organization, in the light of our own experience in the development of the RIVAGE project at Orstom, which aims at simulating runo and inltration processes. We believe that the development of our methods in the eld of physical processes will provide new ideas and tools useful for many multi-agent architectures and mod-elling purposes, so as to give shape to the concept of virtual experiment laboratories.