A Framework to Improve the Disaster Response Through a Knowledge-Based Multi-Agent System

1. Introduction

According to Magdelaine (2017), an average of 413 natural disasters per year are happening during the period of 2000 to 2016, against an average of 256 natural disasters per year for the time period 1983-1999. With this increase of disaster number in the world, a majority of countries in the world is prepared to face disasters. Their preparation is composed mainly of the plan design to define roles and responsibilities of organizations involving in disaster response and required resources according to the level of disaster gravity. Some of the organizations have their own plan of action to achieve their role (also called standard operation) and some others make-decision and acts according to their knowledge of the managed situation. An efficient response can be obtained only if these different organizations collaborate. The two aspects of the collaboration (coordination and cooperation) still face some problems. The aspect of coordination is generally managed by the plan design. However, drills to test the efficiency of these plans have a high cost and are not always applicable to large scale. That is why the simulation systems are increasingly used to test strategies or action plan and assess their efficiency. Concerning the cooperation aspect, problems intervene mainly in the information sharing. A research work for improving the efficiency of response has identified problems of communication and information sharing intervenes at different levels: Community, Agency and Individual (Bharosa, Lee, & Janssen, 2010). Moreover, Manoj and Baker (2007) claim that “Sharing and dissemination of information are both crucial and problematic…” (p. 52). Bharosa et al. (2010) claim that “At the national level, authorities and laws can stimulate inter-organizational information sharing by implementing institutional mechanisms that dictate organizational policies and guidelines” (as cited in Goes & Park, 1997, p. 61). However, the creation of such organizational policies and guidelines requires the understanding and the identification of “good practices” in the information sharing to improve the efficiency of response. This paper is an extension of a previous work (Prudhomme, Roxin, Cruz, & Boochs, 2017) and aims at presenting a framework providing a solution for these two problems of collaboration by (1) simulating the cooperation of the different organizations involving in disaster response to identify potential problems, (2) testing different strategies of information exchange to identify their impact on the response efficiency. The framework’s purpose is to facilitate the simulation of organization collaboration through a generic multi-agent system able to adapt its work according to the knowledge representation given in input. The second section of this paper describes the disaster response and the collaboration. The third section gives an overview of work related to multi-agent systems and more precisely related to agent decision-making and agent collaboration in multi-agent simulations for disaster response. It also provides an overview of ontologies use in disaster response and in multi-agent simulation. The fourth section presents the components of the framework and details the multi-agent system with agent decision-making based on knowledge process and a system of task decomposition.