Ψηφιακό δίδυμο για την έξυπνη διαχείριση απορριμμάτων με χρήση τεχνητής νοημοσύνης

Σαγιά, Αναστασία

Digital twin for smart waste management using Artificial Intelligence

Master Thesis

Author

Σαγιά, Αναστασία

Date

2026

Abstract

Municipal waste collection remains one of the most resource-intensive public services in urban environments. Traditional fixed-route systems operate on static schedules that do not reflect real-time waste generation patterns, leading to unnecessary travel, suboptimal vehicle utilization, avoidable fuel consumption, and increased greenhouse gas emissions. This study develops and evaluates a digital twin framework for smart waste management supported by artificial intelligence-driven forecasting and dynamic routing optimization. The proposed system integrates bin-level waste accumulation modeling, machine learning-based fullness prediction, and dynamic capacitated vehicle routing within a unified digital twin architecture. Waste generation is modeled as a time-dependent stochastic process incorporating weekly seasonality, holiday effects, and contextual variability. Predictive models, including ensemble learning and sequence-based approaches, estimate time-to-threshold for each bin. Routing decisions are recalculated daily using heuristic optimization methods to minimize total travel distance while respecting vehicle capacity and operational constraints. A comparative simulation between a baseline fixed-route system and the AI-optimized digital twin scenario was conducted for a city network of 500 bins and 8 diesel vehicles. Results demonstrate a 26.5 percent reduction in total route distance, a 31.9 percent reduction in fuel consumption, and a corresponding 31.9 percent decrease in CO2 emissions. Vehicle utilization improved from 63 percent to 84 percent, while overflow incidents were reduced from 47 to 9 events per month. Sensitivity analysis across small, medium, and large city networks confirms that optimization benefits scale positively with network size. The findings indicate that operational optimization through digital twin integration provides immediate economic and environmental benefits without requiring fleet electrification or major infrastructure replacement. While implementation challenges such as sensor reliability, data integration, and organizational adaptation must be addressed, the results demonstrate that combining real-time monitoring, predictive analytics, and dynamic routing offers a practical pathway toward more sustainable and resilient urban waste management systems.

Postgraduate Studies Programme

Κλιματική Κρίση και Τεχνολογίες Πληροφορικής και Επικοινωνιών / MSc in Climate Crisis and Information and Communication Technologies

Department

Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών Συστημάτων

Number of pages

102

Language

Greek

URI

https://dione.lib.unipi.gr/xmlui/handle/unipi/19314

Collections

Τμήμα Ψηφιακών Συστημάτων

Show full item record

Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα