Agent-oriented modeling and development of a system for crisis management

• A multi-agent-based approach for crisis management with a novel interaction organization pattern.
• Emergent organization of peers according to circumstances with an efficient use of messages.
• Detailed comparison with the most similar work, the RoboCup Rescue project.
• Implementation of a crisis management system for the scenario of a poisonous material spread in a city.
• Development of tests for validating the crisis management system and showing its performance.

The management of crisis situations has been a challenging problem from different points of views, such as communication efficiency and avoiding casualties. This paper presents a novel approach that includes an interaction organization pattern for Multi-agent Systems (MASs) in crisis management, abstracted from several existing case studies in which the agents follow a sequence of interactions and the organization must optimize the use of human resources. The pattern considers an emergent organization of peers that adopt different roles according to the circumstances. The key features of the organization are its robustness, scalability (in terms of both agents and roles), flexibility to deal with a changing environment, and the efficient use of resources. In order to validate the organization, the paper presents its modeling and development with the Ingenias methodology, conforming the corresponding MAS. This development follows a model-driven approach, which allows a smooth transition from the specification to the code, and a low-cost testing of the system with different settings. Another key aspect is the application of metrics for validating and improving the MAS in terms of response time. The MAS has been tested with 600 agents representing 200 citizens, showing its performance.