AAMAS Conference 2010 Conference Paper
An Architecture for Modular Distributed Simulation with Agent-Based Models
- David Scerri
- Sarah Hickmott
- Alexis Drogoul
- Lin Padgham
Agent-based simulations are an increasingly popular meansof exploring and understanding complex social systems. Inorder to be useful, these simulations must capture a range ofaspects of the modeled situation, each possibly requiring distinct expertise. Moreover, different paradigms may be usefulin modelling, ranging from those that use many lightweightreactive agents, to those that use cognitive agents, to thosethat focus on agent teams and organisational structures. There is need for an architecture which supports the development of a large simulation, through the integrationof separately developed modules. This paper describes aframework and architecture which facilitates the integrationof multiple agent-based simulations into a single globalsimulation. This architecture naturally supports distributedsimulation and incremental development, which are waysof addressing the computational and conceptual complexityof such systems. In this paper we focus particularly onhow to ensure proper management of simulation data thatis affected by agents in different modules, at the samelogical time. We also provide some preliminary performanceevaluation addressing scalability, as well as a comparison ofhow other available systems handle the issue of shared data.