Abstract—Disaster management is a complex collaborative process involving several stakeholders from different domains and requiring preparation for designing action plans. Computer simulation of such collaborative process allows globally assessing the efficiency of such preparation. Studies have shown that multi-agents systems (MAS) are well suited for identifying an optimal strategy or potential issues in the context of one or several action plans. Thus, our approach relies on a MAS for simulating action plans. We consider combining such an approach with semantic Web technologies, in order to define the conceptual model of the simulation according to the preparation model of disaster management. The knowledge base is expressed through two different ontologies (semDM to model the preparation results and semMAS to model the simulation) that will be discussed in this paper. On top of these knowledge models, domain-specific constraints allow for checking consistency and logic rules are used to define the semMAS modeling according to semDM. Index Terms—Disaster management, multi-agent simulation, semantic web technologies, modeling. I. INTRODUCTION Disaster management comprises four steps: Mitigation, Preparedness, Response and Recovery [1]. This cycle of four steps aims at improving the critical step of Response from the analysis of risks and past experiences. Response step requires the most efficiency. However, the constraints of time and the stress during this step limit its efficiency. That is why the key step of the cycle is the Preparedness which allows stakeholders to get ready by organizing, preparing and training to respond efficiently to a disaster. The preparation is done both collaboratively through the distribution of responsibilities between the different stakeholders (e.g. communal plan in France [2]), and individually by each stakeholder through plan and protocol design according to their business knowledge (e.g. firefighter protocol [3]). However, during the response, all the stakeholders have to work collaboratively to face the disaster and actions of one can impact the other. That is why the application of the individual preparation produces some problems or a loss of efficiency. Manuscript received December 15, 2018; revised May 1, 2019. This work was supported by the German Federal Ministry of Education and Research (“Bundesministerium für Bildung und Forschung”) [grant numbers 03FH032IX4]. Claire Prudhomme and Frank Boochs are with the institute i3mainz of the University of Applied Sciences, Lucy-Hillebrand-Str. 2, 55128 Mainz, Germany (e-mail: claire.prudhomme@hs-mainz.de, frank.boochs@hs- mainz.de). Christophe Cruz is with the Laboratoire d'Informatique de Bourgogne (LIB) - EA 7534, University of Bourgogne Franche-Comté, Bâtiment i3M rue Sully, 21000 Dijon, France (e-mail: christophe.cruz@ubfc.fr). To gain in experience without waiting a new disaster, drills are organized. Drills are a training that aims at assessing the preparation of the different response stakeholders sometimes including the population. However, the high cost of drills limits their quantity and their quality. The disaster management faces to a lack of collaborative and shared point of view during the preparation that produces a loss of efficiency to respond a disaster. This lack of shared point of view due to the low number of collaborative drills must be fulfilled by another way. Such as shown by drills, a shared feedback on experience requires the application of collaborative and individual preparation in a situation of disaster to assess its result with a shared point of view. The collaborative and individual preparation makes intervene a set of organizations composed of different actors with different roles and responsibilities, a set of plans (distribution of task responsibilities), a set of protocol (set of actions and resources required to achieve a task). Among the different business domain intervening in the disaster management, some have predefined protocols and some others use their business knowledge to build protocols according to the situation of disaster. A situation of disaster is composed of a set of events that impact the population and the infrastructures in the considered geographical area. Studies have shown that multi-agents simulation (MAS) are well suited for identifying an optimal strategy or potential issues in the context of one or several action plans. The level of granularity provides by a multi-agent simulation allow for representing the diversity of organizational structure, actors and behaviors required to represent the preparation result. However, the existing works have a conceptual model designed for a specific use case or propose metamodel allowing for designing a diversity of use cases. But, they do not provide approach to allow the generation of a conceptual model from an explicit model of preparation. A conceptual model for a realist simulation of preparation results has two prime input requirements. Firstly, geospatial data containing the location and information about the population and infrastructures must be gathered and linked to information related to disaster situation and its impacts. Secondly, information and knowledge about organizational structure, actors, tasks, actions and resource derived from the different stakeholders must also be related to the disaster management. According to the location, the type and the gravity of the disaster situation, the preparation model associated varies to fit with the requirements of the situation. To simulate preparation result, its conceptual and implemented models have to be adapted according to various possible disaster situations and thus, various preparation models. The previously cited data and information allow for configuring the preparation model Semantic and Logic Modeling of Disaster Simulation for Multi-agent Systems Claire Prudhomme, Christophe Cruz, and Frank Boochs International Journal of Modeling and Optimization, Vol. 9, No. 4, August 2019 198 DOI: 10.7763/IJMO.2019.V9.709