emergency transport optimized by Artificial Intelligence, automatic natural language processing and deep learning – SADIAND

The SADIAND joint laboratory is part of a context that favors the development of artificial intelligence, with increasingly large volumes of available data, ever more powerful algorithms and machines, and greater complexity in decision-making due to an uncertain and changing environment. Its application within Civil Security, for the benefit of populations and their protection, is a major challenge.

SADIAND proposes to establish a perennial collaboration between FEMTO-ST and SAD MARKETING to develop a new generation of intelligent solutions for the fire and rescue sector, with a double objective: flow forecasting and optimization of the operational response. The ultimate goal is to guarantee the continuity of public emergency services.

SADIAND combines the expertise of SAD MARKETING and FEMTO-ST.

- SAD MARKETING is a French SME that has been working for 40 years on the French and international markets. The company historically accompanies major players in the retail industry (Carrefour, Decathlon...) on the establishment of new points of sale through spatialized sales forecasting. For the past year, SAD MARKETING has expanded its areas of intervention to the fire and rescue sector by deploying a predictive analysis solution for SIS. This opening on this new market is the result of a collaboration with FEMTO-ST which had initiated this research work with the SDIS 25. SAD MARKETING wishes to accelerate on this market by making these models more reliable and by a wider coverage allowing to cover all the risks faced by the rescue actors (forest or crop fires, transport accidents, domestic accidents, etc.).

- The FEMTO-ST Institute, through its transverse axis, is an expert in artificial intelligence. The work of the laboratory concerns various subjects ranging from the prediction of anomalies or defects, to quality control on images, to the segmentation of cardiac MRI, but also around the automatic processing of natural language. The laboratory has expertise on different types of data: 1D with time series or text; 2D with images; or 3D with a set of images. The laboratory has developed several collaborative projects with companies through CIFRE theses. Moreover, the laboratory is involved in several ANR projects around the themes of artificial intelligence.

Steps of the project are as follows:

In the short term, the objective is to improve existing predictive models by adding new explanatory variables from automatic natural language processing. Several potential data sources have already been identified, although these are not limited: weather reports, social networks (twitter, Snapchat...).

In the medium term, the objective is to improve risk coverage by processing satellite images. Several risks can be addressed by image processing, in particular forest fires or fires related to agriculture. In the medium term, we also want to integrate population mobility data from GPS data from mobile applications or telephone operators, as denser areas reflect an increased risk.

In the long term, we want to work on optimization algorithms in order to propose the most appropriate operational response. The mathematical models will also be adapted to new markets: international, hospital SMUR, and private ambulances.