Skip to Main content Skip to Navigation

Distribution et Parallélisation de Simulations Orientées Agents

Abstract : Simulation of complex systems provides observation of the studied systems' dynamics in a suitable time for scientific interpretation. Agent-oriented simulation features multi-scales observation by modelling the system's granular entities and their interactions. However appealing, execution of large-scale simulations implying throngs of agents is a major issue. This work aims at executing such system on a network of simulation platforms. After having identified requirements of this approach, we propose an agent-based architecture to manage the consistency of distributed simulation on a flexible execution infrastructure. The multiagent system composed of agents distributed among the infrastructure establishes a virtual platform that executes simulation without requiring adaptation of the simulated model. In order to improve execution performances, we consider time representation and agents interactions, more precisely the perception/influence model, enabling parallel scheduling of the simulation. Along with this scheduling, we propose a dynamic load-balancing algorithm. This one is also based on simulation model's information and aims at both maximising exploitation of each platform and ensuring the flexibility of simulation regarding the flexibility of the execution infrastructure. Concepts and algorithms we propose are then implemented in our platform GEAMAS-NG.
Document type :
Complete list of metadatas

Cited literature [10 references]  Display  Hide  Download
Contributor : Nicolas Sébastien <>
Submitted on : Monday, April 19, 2010 - 12:57:50 PM
Last modification on : Monday, September 2, 2019 - 9:42:45 AM
Long-term archiving on: : Monday, October 22, 2012 - 3:10:22 PM


  • HAL Id : tel-00474213, version 1


Nicolas Sébastien. Distribution et Parallélisation de Simulations Orientées Agents. Autre [cs.OH]. Université de la Réunion, 2010. Français. ⟨tel-00474213⟩



Record views


Files downloads