Vittorio Alberto Gianfrate
MARES Madrid Emergency Response System: A Multi-Agent Reinforcement Learning Architecture for Real-Time Meteorological Crisis Response.
Rel. Lia Morra. Politecnico di Torino, Master of science program in Computer Engineering, 2026
Abstract
The growing frequency and severity of meteorological hazards has exposed fundamental weaknesses in conventional emergency management frameworks. Operators face cognitive overload from persistent low-priority notifications, decision thresholds remain static despite changing environmental conditions, and the tension between the speed advantages of autonomous systems and the legal requirements for human accountability remains unresolved. This thesis presents the Madrid Emergency Response System (MARES), a protocol-driven Multi-Agent System (MAS) for real-time meteorological crisis management within the Community of Madrid, implemented as a mobile application operable from a smartphone or tablet by civil protection operators in the field. The architecture coordinates fourteen specialised asynchronous agents over an event-driven publish-subscribe bus, incorporating live meteorological data ingestion from AEMET OpenData, territorial vulnerability assessment, and full forensic traceability through an immutable append-only relational persistence layer.
Three architectural mechanisms directly address documented failure modes in existing decision-support systems
Relators
Academic year
Publication type
Number of Pages
Additional Information
Course of studies
Classe di laurea
Ente in cotutela
Aziende collaboratrici
URI
![]() |
Modify record (reserved for operators) |
