Here you can find several publications made by the people who are working or worked on MECSYCO
Multi-agent approach has demonstrated its benefits for complex system modeling and simulation. This article focuses on how to represent and simulate a system as a set of several interacting simulators, with a focus on the case of multi-agent simulators. This raises a major challenge: multi-agent simulators are not conceived (in general) to be used with other simulators. This article presents a preliminary study about the rigorous integration of multi-agent simulators into a co-simulation platform. The work is grounded on the Net-Logo simulator and the co-simulation platform mecsyco.
Study about decomposition and integration of continuous systems in discrete environment
| Thomas Paris, Alexandre Tan, Vincent Chevrier, Laurent Ciarletta
A complex system is one composed of many interacting heterogeneous entities. This kind of system can be dealt with multi-modeling and co-simulation but individual models may also be heterogeneous (continuous , discrete, event-based...). To manage this complexity , we use MECSYCO (Multi-agent Environment for Complex-SYstem CO-simulation) a DEVS compliant environment for co-simulation. MECSYCO handles heterogeneity issues, but the number of models which may interact during a co-simulation of a complex system raises also performance issues. So it's important to develop performance measurement tools to study MECSYCO's co-simulation performances. In this article we present modular performance measurement tools for MECSYCO. We test these tools on our " Multi-Room Heating " model, a scalable continuous system, to assert the tradeoff between accuracy and computational time when integrating continuous system in a discrete modeling environment. Then we study the impact of decomposing a continuous system contained in one FMU into several FMUs which interact. We verify the validity of our tools and we show that, under some conditions, a large model that cannot be solved on one block, can be decomposed into smaller ones, solved and simulated in a co-simulation on MECSYCO without significant loss of accuracy.
Multi-agent multi-model simulation of smart grids in the MS4SG project
| Julien Vaubourg, Yannick Presse, Benjamin Camus, Christine Bourjot, Laurent Ciarletta, Vincent Chevrier, Jean-Philippe Tavella, Hugo Morais
This paper illustrates how the multi-agent approach, or paradigm, can help in the modeling and the simulation of smart grids in the context of MS4SG (a joint project between LORIA-INRIA and EDF R&D). Smart grids simulations need to integrate together pre-existing and heterogeneous models and their simulation software; for example modeling tools of the power grids, of telecommunication networks, and of the information and decision systems. This paper describes the use of MECSYCO as a valid approach to integrate these heterogeneous models in a multi-agent smart grid simulation platform. Several use cases show the ability of MECSYCO to effectively take into account the requirements of smart grids simulation in MS4SG
We are interested in the modeling and simulation of the IP networks as the communication layer for smart grids. In this context, a network can involve many different technologies, and the available models corresponding to these technologies may be implemented in different simulation software. As a result, thanks to their own models libraries, different IP network simulators can be complementary. However, these simulators are not all interoperable with each other, and therefore cannot be yet all integrated in a same co-simulation. Moreover, the network simulators have to interact with the other simulators corresponding to other areas of expertise involved in the smart grids simulations. Integrating this requires to consider the multi-formalism problems. Our approach is to integrate IP network simulators to the DEVS-based co-simulation platform MECSYCO (formally named AA4MM). Thanks to this approach, different network simulators can exchange simulated IP packets. In this paper, we illustrate how we integrated the NS-3 simulator into MECSYCO.
We are interested in the multi-modeling and simulation of complex systems, that is representing a complex system as a set of interacting models and simulating it with a co-simulation approach. Representing and simulating a complex system multi-model requires to integrate heterogeneity at several levels (representations, formalisms, simulation software, models' interactions.. .). In this article, we present our approach that consists of combining the Discrete EVent System Specification (DEVS) formalism and multi-agent concepts in order to achieve these requirements. The use of the DEVS formalism enables a rigourous integration of models described with heterogeneous formalisms and a rigourous simulation protocol. Multi-agent concepts ease the description of multi-perspective integration and the reuse of existing heterogeneous simulators. We detail the combination of both in the Agent & Artifact for Multi-Modeling (AA4MM) approach and illustrate its use in a proof of concept.
We propose to consider a multi-level representation from a multi-modeling point of view. We define a framework to better specify the concepts used in multi-level modeling and their relationships. This framework is implemented through the AA4MM meta-model, which benefits from a middleware layer. This meta-model uses the multi-agent paradigm to consider a multi-model as a society of interacting models. We extend this meta-model to consider multi-level modeling and present a proof of concept of a collective motion example where we show the ability of this approach to rapidly change from one pattern of interaction to another one by reusing some of the meta-model's components.
Modélisation multi-niveaux dans AA4MM
| Benjamin Camus, Julien Siebert, Christine Bourjot, Vincent Chevrier
In this article, we propose to represent a multi-level phenomenon as a set of interacting models. This perspective makes the levels of representation and their relationships explicit. To deal with coherence, causality and coordination issues between models, we rely on AA4MM, a metamodel dedicated to such a representation. We illustrate our proposal and we show the interest of our approach on a flocking phenomenon.
Impact des dimensions spatiale et temporelle dans la modélisation d'un phénomène collectif de type free-riding
| Tomas Navarrete Gutierrez, Julien Siebert, Laurent Ciarletta, Vincent Chevrier
We present a comparison of five different models built upon the same individual behavior hypothesis of a collective phenomenon present in peer-to-peer file exchange networks: "free-riding". We study a global analytical model and four agent based models. Multi-agent models include the space and time dimensions rarely seen in the literature discussing aggregated models of the collective phenomenon in question. We discuss the a priori and the experimental conditions under which the models are equivalent. We demonstrate that one individual decision algorithm can lead to contradictory information.
Complex systems simulation implies the interaction of different scientific fields. However, most of the time people involved into the simulation process do not know intricate distributed simulation tools and only care about their own domain modelling. We propose a framework (called AA4MM) to build a simulation as a society of interacting models. The main goal is to reuse existing models and simulators and to make them interact. The coordination challenges remain to the AA4MM framework so that the simulation design and implementation stay as simple as possible. In this paper, we present the coordination model which intends to decentralize the simulators interactions. We propose to use the environment through the notion of artefact in order to deal with the coherence, compatibility and coordination issues that appear in parallel simulations.
AA4MM coordination model and event-B specification
| Julien Siebert, Joris Rehm, Vincent Chevrier, Laurent Ciarletta, Dominique Méry
We develop a framework called Agent and Artefact for Multiple Models coordination (AA4MM) . It is is intended to make the design and the implementation of complex systems simulation modular and decentralized. Our main goal is to reuse existing models and simulators and to make them interact in order to simulate dierent levels of abstraction. The main constraint is that people involved into the design process do not have to care about anything else but modelling. Coordination challenges remain to the framework. This report presents the event-B specication of the coordination model proposed in [SCC10]. Its goal is to check that the system is never blocked (no deadlock): no simulator is waiting for another indenitely.
Cet article s'inscrit dans le cadre de l'étude des systèmes complexes via la modélisation et la simulation informatique. Nous pensons qu'il est parfois nécessaire de faire interagir plusieurs modèles pour simuler un phénomène. Dans le cas des technologies des réseaux dynamiques (réseaux P2P, réseaux mobiles Ad Hoc), le comportement des usagers et le fonctionnement des réseaux s'inﬂuencent mutuellement. Nous proposons une approche de modélisation et un outil de simulation couplant un modèle d'utilisateurs - basé sur le paradigme multi-agents - et un modèle de réseaux dynamiques. Nous discutons, au travers d'un cas d'étude particulier, des avantages, des problématiques soulevées et des limitations d'une telle approche. Nous montrons que cette démarche de modélisation apporte un niveau de précision et une ﬂexibilité élevés.
Dans cet article, nous présentons les interconnexions qui peuvent se faire entre, d'une part, la modélisation et simulation multi-agents, et le domaine des réseaux pair-à-pair (P2P) d'autre part. Plus particulièrement, dans ce domaine, nous traitons de la prise en compte, dans les modèles de simulation, des comportements de l'utilisateur sur l'une des caractéristiques du fonctionnement de ces réseaux : la qualité de service. Dans un premier temps, nous montrons les caractéristiques qui nous semblent souhaitables pour modéliser de manière adéquate de tels réseaux, montrons les limites actuelles que nous constatons dans le domaine multi-agents et proposons une approche multi-modèles. Dans un second temps, nous détaillons la mise en oeuvre de cette proposition puis validons son bien-fondé en détaillant quelques expérimentations réalisées qui montrent que notre outil correspond bien à nos attentes de modélisation.
In distributed, dynamic networks and applications, such as Peer-to-Peer (P2P), users' behaviour and quality of service/quality of experiment are known to influence each other. In worst cases, these mutual influences could lead the system to crash. We propose a novel approach to model relationships between users and QoS. It is based upon multi-agent systems in order to study the impact of situated behaviours on the global network and to integrate different levels of representation (users' behaviour, overlay protocols, network topology). This paper describes our approach to represent the different models required in such systems and a first implementation in an existing overlay simulator with the first results of experimentations.
Dans cet article, nous présentons une démarche conceptuelle liant des modélisations issues des mondes pair-à-pair et multi-agents afin de prendre en compte le comportement de l'utilisateur dans la modélisation et la simulation des réseaux pair-à-pair. Nous présentons aussi un outil de simulation que nous avons adapté ainsi que nos premiers résultats qui indiquent que cette démarche est cohérente.
Network access and services are becoming ubiquitous, and the number of their users and usage is still growing rapidily. Controlling those networks is incresaingly complex. At the same time, the notion of infrastructure is also shaken by new technologies such as P2P or adhoc networks. Standard control and evaluation mechanism are not taking into account the complexity, diversity and dynamicity of the users' behavior, which are the subject of study of multi-agent simulation. This document explores the opportunity to bridge the usual networking modelling and simulation tools with the multi-agent approach.