INTEGRISANI MODEL LOKALIZACIJE SAOBRAĆAJNIH ENTITETA ZASNOVAN NA DUBOKOM UČENJU I MIKROSIMULACIJI KAO ALAT ZA UNAPREĐENJE ODRŽIVE URBANE MOBILNOSTI / INTEGRATED TRANSPORT ENTITY LOCATION MODEL BASED ON DEEP LEARNING AND MICROSIMULATION AS A TOOL FOR IMPROVING SUSTAINABLE URBAN MOBILITY

Abstract

Modern urban transportation systems face challenges such as congestion, increased greenhouse gas emissions, and reduced energy efficiency, which require data-driven solutions in mobility planning. This study investigates the contribution of an integrated traffic entity localization model, based on deep learning and microsimulation, to improving urban mobility sustainability. The proposed approach integrates traffic entity detection from video recordings using the YOLOv8 model, spatial analysis through GIS tools, and traffic flow microsimulation in the SUMO environment. The model includes a physical traffic network model, an analog model of entity behavior, and a mathematical model for evaluating urban mobility sustainability indicators. Empirical validation was conducted on selected urban intersections in Sarajevo Canton using real video and infrastructure data. The results confirm that integrating deep learning and microsimulation provides an effective tool for analysis and decision support in sustainable urban transport planning.