A Digital Twin–Driven Predictive Traffic Management Framework Incorporating Driver Micro-Behavior Modeling

Abstract

Modern intelligent transportation systems (ITS) rely heavily on reactive control strategies that respond only after congestion emerges. This paper proposes a Digital Twin–Driven Predictive Traffic Management Framework, enhanced with a Driver Micro-Behaviour Modelling (DMBM) module, to forecast short-term traffic evolution and apply proactive interventions. The digital twin synchronizes with real-world traffic every 1–5 seconds and uses a hybrid LSTM and Graph Neural Network (GNN) model to predict traffic states up to 30–300 seconds ahead. The DMBM module introduces fine-grained behavioural parameters—aggression index, compliance probability, reaction time distribution, vehicle familiarity, and gap acceptance thresholds—enabling realistic micro-simulation dynamics. Interventions (signal timing optimization, lane reassignment, dynamic speed regulation, and pre-emptive routing) are tested within the digital twin before deployment. Experiments conducted using Simulation of Urban Mobility (SUMO) show improvements of 18–42% in congestion prevention, 15–28% reduction in travel time, and 20–37% faster emergency response routing compared with leading reactive adaptive traffic systems. Results confirm that integrating micro-behaviour modelling within a real-time digital twin provides a significantly more accurate and proactive traffic management solution.