MODELING, CALIBRATING, AND PREDICTING TRAFFIC JAM LENGTHS UNDER MIXED AGGRESSIVE AVS AND HUMAN DRIVING SCENARIOS

Autonomous Vehicles, Aggressive Driving Behavior, Traffic Congestion, Signalized Intersection, Regression Modeling. This study evaluates the impact of varying penetration rates of aggressive autonomous vehicles (AVs) on traffic congestion at a four-legged signalized intersection in Al-Qadisyah, Kirkuk City, Iraq. A simulation using SUMO and regression modeling quantified the relationship between penetration rates and congestion metrics. Introducing 25% aggressive AVs reduced the Mean Max Jam Length by approximately 7% compared to 100% human-driven scenarios, while a full transition to 100% aggressive AVs achieved a 30% reduction. Linear and quadratic regression models were calibrated to represent these relationships, with quadratic models demonstrating superior accuracy based on validation metrics like Coefficient of determination. (R²), Mean Squared Error (MSE), and Mean Absolute Error (MAE). Predictions for intermediate penetration rates aligned with observed trends, offering actionable insights for traffic planning. This research highlights the potential of aggressive AV behavior in reducing congestion, optimizing intersection performance, and informing AV deployment strategies, supported by rigorous calibration and validation of predictive models.