Radar-Based Evaluation of Car-Following Behavior and Fuel Consumption Across Vehicle Categories in SUMO

Abstract

This study evaluates how well the Intelligent Driver Model (IDM) reproduces observed leader--follower behavior across vehicle categories and how these behavioral differences influence PHEMlight fuel-consumption estimates in SUMO. The analysis uses fused roadside radar trajectories collected along a 1.5 km signalized arterial corridor and classifies follower vehicles into four categories using radar-reported length: Cars, SUVs, light-duty Trucks, and Class-8 vehicles. Measured leader trajectories are replayed in SUMO via TraCI, while followers are simulated using IDM. Default parameters and parameters calibrated with a trajectory-level objective in spacing, velocity, and acceleration are compared using RMSE in spacing, velocity, and acceleration. Calibration reduces RMSE across all categories, with the largest reductions in spacing, but Class-8 trajectories retain higher reproduction error than the light-duty categories. A focused Class-8 analysis shows that erroneously simulating the same Class-8 leader--follower trajectories as vClass passenger increases uncalibrated spacing and acceleration error relative to the Class-8 (vClass trailer) representation, while calibration eliminates the difference between the two representations. PHEMlight fuel-consumption medians decrease after calibration for each vehicle category, with the largest absolute reduction observed for Class-8 vehicles. These results highlight the importance of category-aware behavioral modeling when microsimulation outputs are used for energy and emissions assessment in mixed traffic.