Preventing flood risk with decision algorithms

Within the framework of the Criz’Innov and i-Nondations, the teams SEPIA (Operating System, Distributed Systems, from Software to Architecture) and SMAC (Cooperative Multi-Agent Systems) present, in video, how decision algorithms can help anticipate and manage a crisis. Using the GAMA multi-agent platform, researchers worked on two simulations of flooding situations in the city of Trèbes.

Preventing the risk of flooding

The consortium of the ANR i-Nondations project makes the following observation :

“Every year, floods occur. Solutions exist for slow floods, but rapid floods are difficult to predict and manage.”

This innovative scientific project, started in 2018 and ending in August 2022, aims to model rapid flooding. The data collected by technological or human sensors allows to anticipate the risks and optimize the impact of a flood on the infrastructures. It proposes to manage three phases which are the periods before, during and after a crisis, in a feedback loop coming from the autonomy domain called MAPE-K loop. It is based on four stages: Monitoring, Analysis, Planning and Execution with a knowledge database. Monitoring detects events, analysis analyzes risks based on predictive models, creates metrics to measure the impact on the infrastructure and proposes functions to evaluate alternative trajectories. Planning defines how the crisis should be managed, among several possible scenarios. Execution is the phase of putting the chosen planning scenario into action.

Managing a crisis situation

GAMA is used to simulate evacuation strategies in the city in a crisis situation, that of a flood. GAMA is a modeling and simulation development environment for building spatially explicit agent-based simulations. GAMA has been developed with a very general approach and can be used for many application domains. Some additional plugins have been developed to meet specific needs. The application domains where GAMA is most present are transportation, urban planning, epidemiology and environment. In the framework of the Criz’Innov project, GAMA was used to simulate two flooding scenarios in the French city of Trèbes. The first simulation uses the algorithms developed in the i-Floods project. The latter has two main objectives: to estimate the total evacuation time and to determine the number of buses needed in order to avoid using resources that are more difficult to deploy. These simulations are performed using geo-spatial data. These data correspond, for example, to the location of priority issues to be evacuated in case of flooding, to the roads that can be used or to the calculation of the seven levels of flooding. This tool also includes an optimization algorithm that calculates the best route to take for each vehicle. All these data will allow the optimization of the use of vehicles according to the different possible flood levels.

Protecting people

The second simulation was carried out to optimize the sheltering of the population, with or without the police. The objective is to estimate the time needed for the population to take refuge in their homes or in evacuation centers. The model can also be used to determine the best location of the gendarmes in the city. All of this data can be used to make comparisons that can be used to adapt the strategy for bringing people to safety. For example, when there is a deployment of gendarmes, the sheltering time is reduced. When we simulate a limitation of movements to only those directed towards a shelter and coordinated by gendarmes, we observe an acceleration of the time to shelter.

These two simulations and the Criz’Innov and i-Nondations projects show how decision tools and algorithms can play a decisive role in crisis or natural disaster situations. They can be used to anticipate a crisis and manage the means and resources available to optimize response strategies. Behaviors can be simulated to better respond to the safety of people.