Where Should Traffic Sensors Go? Sensor Placement for Urban Flow Estimation

Summary

Beschreibung • Traffic sensors observe only a small part of an urban road network. Their locations determine how much these measurements reveal about traffic on unobserved roads. • This thesis investigates where a limited number of sensors should be placed to improve network- wide traffic-flow estimation, especially when historical measurements are unavailable. • Available resources include OpenStreetMap road graphs, synthetic traffic flows, and an existing pretrained graph neural network—a model that learns relationships between connected road segments. • The core study uses synthetic demand scenarios with known flows on every road segment. It compares placement strategies with a fixed estimator and equal sensor budgets; real-data evaluation is an optional extension. Work Packages • Review sensor-placement methods and reproduce random and graph-based baselines • Develop a placement strategy using synthetic flows or information from the existing model • Compare strategies across sensor budgets and road networks on held-out demand scenarios; keep test flows hidden during sensor selection • Analyse estimation errors on unobserved roads, robustness, and computational cost; document reproducible results and limitations Voraussetzungen Prerequisites • Programming skills in Python and basic knowledge of machine learning • Interest in transport networks, graph methods, or optimisation • Independent, structured work and critical interpretation of results