Um framework integrado para resposta rápida a eventos de crise: realocação de transportes contextualizada no sistema ICARUS

Abstract

Extreme climate events, particularly floods, pose complex challenges to urban management, disproportionately affecting vulnerable populations. An effective response to such crises requires integration between detection, logistics, and communication. This dissertation, presented as a collection of articles, proposes an integrated framework for the ICARUS system to optimize flood response. The research is structured around three complementary axes: (1) the development of a Data Science methodology to estimate social vulnerability indices in urban grids using public data; (2) the implementation of graph-based algorithms for public transport rerouting, prioritizing routes that serve the high social risk areas; and (3) the creation of a conversational agent based on Large Language Models (LLMs) to provide real-time, contextualized safety guidance to the population. Results indicate that the incorporation of vulnerability criteria significantly alters mobility strategies without compromising computational efficiency, while the intelligent agent demonstrates effectiveness in disseminating critical information. This work contributes to the Smart Cities literature by demonstrating how different technological layers can operate in synergy to mitigate the impacts of natural disasters.