Do Indian Drivers Conform to Newells Car Following Model?

Abstract

Newell's simplified car-following model is known for its parsimony and behavioural interpretability, but the effects of finite interaction boundaries on parameter estimation remain poorly understood, particularly in lane-free environments. This study tests the model against lane-free Indian traffic using two independent methods: aggregate and pair-specific regression of the speed-spacing relationship, and a shifting optimization that estimates the optimal parameters by minimizing spacing variance. A boundary-corrected variant of the shifting method is also proposed to isolate transitional regimes at interaction endpoints, while preserving the original Newell formulation. The analysis uses high-resolution vehicle trajectories from a highway corridor in Chennai. Pair-specific regression models substantially outperform aggregate specifications, underscoring the role of driver heterogeneity. Trajectory-shifting recovered parameter distributions statistically equivalent to regression estimates, independently supporting the trajectory-translation principle. The boundary correction improved the mean trajectory fit from average $\bar{R^2}$ = 0.66 to 0.95, with the corrected formulation preferred for 91% of pairs. The results point to interaction boundary-effects, not failures in the core behavioural assumptions, as the main source of model error in this setting. Correcting for these effects substantially improves parameter estimation without sacrificing the model's parsimony. These findings establish boundary correction as an essential step in trajectory-based calibration of Newell-type models, with direct implications for traffic-state estimation and microscopic simulation.