Calibration of VISSIM Parameters for Modeling Heterogeneous Traffic Conditions at Intersections in Kathmandu

https://doi.org/10.46610/IJTMTN.2026.v02i01.001

Abstract

Kathmandu’s intersections, characterized by heterogeneous traffic conditions, present unique challenges in traffic modeling. Dominated by motorcycles, these intersections feature diverse vehicle types, exhibit side-by-side vehicle stacking, variable lane widths, and the absence of lane markings and road discipline. In Nepal, microsimulation models are typically developed using trial-and-error methods and are not properly calibrated to such conditions. This study proposes an automated calibration approach for microscopic traffic simulation models, providing calibrated parameter values tailored to local traffic conditions. Three intersections were modeled using VISSIM, a microscopic simulation tool. Sensitivity analysis was conducted using Latin Hypercube Sampling (LHS) and two-level ANOVA testing to identify nine sensitive calibration parameters. Optimization using a bi-level genetic algorithm minimized the error between the simulated and field traffic flow and queue-length data. This resulted in the following sensitive calibration parameters with the recommended range of values—minimum look ahead distance (10−20), look back distance [minimum (15−18.71) and maximum (107.94−150)], average standstill distance (0.3−1.5), safety distance part [additive (0.1−0.5) and multiplicative (0−1)], minimum clearance (front/rear) (0.49−0.73), and minimum lateral distance [standing (0.2−0.41) and driving (0.6−0.9)]. When validated against the videographic survey data, the outputs appear similar to the findings from Indian studies. However, the tolerance levels of the outputs suggest that we need our values rather than Indian values. This methodology and the outputs are expected to significantly reduce the time and effort required for calibrating VISSIM models in similar traffic environments.