From a Multi-agent Simulation Theory to GALATEA

摘要

This paper discusses a simulation theory with learning agentsrnwhich is serving as a formal specification to guide the developmentrnof GALATEA, a multi-agent simulation platform.rnWe have extended an existing simulation language: GLIDER,rnwith abstractions to model systems where autonomous entitiesrn(agents) perceive and act upon their environments. Wernare now applying it to the study of multi-agent systems. Inrnparticular, an implementation on Biocomplexity [1] is brieflyrndiscussed in the paper. We also show how an Inductive LogicrnProgramming system can be used to learn rules in a representationrnvery close to the one used to guide the simulation inrnthe biocomplex system. This establishes the feasibility of embeddingrn(resource-bounded) learners as agents that take partrnin simulating a complex system, as defined by the theory.