Firefighting resource allocation algorithms based on predictions

Ευθύμιος, Ντόκας; Efthymios, Ntokas

dc.contributor.author	Ευθύμιος, Ντόκας	el
dc.contributor.author	Efthymios, Ntokas	en
dc.date.accessioned	2026-06-29T11:36:24Z
dc.date.available	2026-06-29T11:36:24Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/65182
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.32876
dc.rights	Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα	*
dc.rights.uri	http://creativecommons.org/licenses/by-nc-nd/3.0/gr/	*
dc.subject	Χωροθέτηση Εγκαταστάσεων	el
dc.subject	Γεωγραφική Συσταδοποίηση	el
dc.subject	Επιχειρησιακή Έρευνα	el
dc.subject	Πρόβλεψη Κινδύνου Πυρκαγιάς	el
dc.subject	Γραμμικός Προγραμματισμός	el
dc.subject	Facility Location	en
dc.subject	Geographic Clustering	en
dc.subject	Operations Research	en
dc.subject	Linear Programming	en
dc.subject	Fire Risk Prediction	en
dc.title	Firefighting resource allocation algorithms based on predictions	en
heal.type	bachelorThesis
heal.secondaryTitle	Αλγόριθμοι Χωροθέτησης Πυροσβεστικών Πόρων Βασισμένοι Σε Προβλέψεις	el
heal.classification	Algorithms	en
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2026-02-27
heal.abstract	This thesis investigates the problem of firefighting vehicle deployment within a geographical region, integrating spatial data and fire risk predictions. The study addresses both static planning and dynamic operational scenarios. In the static scenario, vehicle locations are optimized to prioritize proximity to high-priority areas, aiming to ensure effective response to potential fire incidents. In the dynamic setting, given an initial deployment, decisions are made regarding which vehicles to dispatch to a new incident and how to reposition the remaining resources to maintain coverage. A comprehensive decision support system is developed, incorporating multi-scale geospatial data, including long-term static data, daily dynamic data, and risk predictions generated through artificial intelligence. Graph-based clustering algorithms are applied to partition the study area into spatially homogeneous regions, enabling balanced allocation of available vehicles. Candidate deployment locations are selected using the Overpass Turbo API, heuristic methods, and a min–max selection scheme to ensure uniform coverage across each cluster. Travel distances are computed based on actual road networks and accessibility metrics through the OpenRouteService API. An importance index is assigned to each candidate location, reflecting the need for nearby vehicle presence. Linear optimization models are formulated for both static and dynamic scenarios, incorporating the importance index, selected candidate locations, and network-based travel distances. Static deployments are evaluated using cumulative distribution functions (CDFs) for k vehicles based on historical fire data, supported by visibility visualizations and comparative analysis of alternative layouts. A case study demonstrates two distinct operational scenarios and compares the resulting performance outcomes. All components are integrated into a web-based application that allows dynamic input of operational data, parameter adjustments, and real-time recalculation of deployments. The full source code is publicly available in a GitHub repository to ensure transparency and reproducibility. Expert knowledge from professional firefighters and domain scientists was used to validate the applicability of all the proposed methodologies.	en
heal.sponsor	Athena Research Center provided funding through a research internship grant.	en
heal.advisorName	Φωτάκης, Δημήτριος	el
heal.committeeMemberName	Φωτάκης, Δημήτριος	el
heal.committeeMemberName	Παγουρτζής, Αριστείδης	el
heal.committeeMemberName	Συμβώνης, Αντώνιος	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών	el
heal.academicPublisherID	ntua
heal.numberOfPages	103 σ.	el
heal.fullTextAvailability	false
heal.fullTextAvailability	false