7.1. Introduction

Implementing multi-agent simulations means designing, building and modeling a complex system made up of individualized entities having a certain degree of autonomy and interacting with each other. Such an approach is proposed today for the resolution of artificial intelligence problems as well as for representing economic, ecological, geographical or physical processes. It is used increasingly in the explanation of spatial dynamics.

The studies based on multi-agent simulations are numerous. We can cite some pioneer applications from the 1990s such as Cinefil in parasitology [PIC 91], SimDelta in halieutics [BOUS 93], Manta in animal ethology [DRO 95], Sealab [LEP 96] in marine ecology and Simpop [BUR 96] in urban geography. All these applications have contributed to establish connections between the disciplines involved and computer science. The examples below will illustrate the diversity of modeled fields. All of them give an important place to the space in which the studied dynamics are deployed.

In the field of physical dynamics, Rivage [SER 00] is a simulator of water runoff processes. Research from Teles [TEL 98] focuses on multi-agent simulation of the alluvial plain genesis. Research from Breton [BRE 00] proposes a multi-agent method to calculate the network of forces in a pile of sand at equilibrium, with a problem resolution perspective (ecoresolution).

In the marine ecosystems and halieutics fields, Ichtyus [MES 94] was programmed for the study of gregarious behaviors and school formation by asking the question: what minimal...