Simulating and Analyzing Crowdsourcing Impacts in Flooding Management: a geo-spatial agent-based approach
Abstract
Crowdsourcing is becoming essential to facilitate disaster management. It can help to gather or check information and delegate collective physical tasks (alerting, rescuing, sheltering, food distribution…) to volunteers. In this context, the management of volunteers, with possible uncertain behavior, cannot be improvised, for both security and efficiency reasons. Therefore, it is necessary to provide officials with tools to anticipate and prepare coordination with volunteers. For that purpose, this paper proposes a geospatial agent-based simulator to visualize, measure and analyze the influence of crowdsourcing in natural disasters management. More precisely, this tool allows the authorities to visualize a crisis situation (actors, environment) and its evolution throughout time and space, improve their situation awareness and explore several what-if scenarios so as to ease coordination with official responders. Moreover, task assignment is implemented according to the contract net protocol to select the volunteers according to their variable characteristics: availabilities, positions, skills… This paper describes the design and implementation of this simulator, which is based on a conceptual model representing the environment, agents’ behaviors and their interactions. We also demonstrate its use through a real-world case study based on a flood that took place in 2018 in Trèbes, a French town. We demonstrate with quantitative indicators the positive impacts of crowdsourcing on this crisis management. This simulator could be easily reused for other natural disaster situations.