Publication: Investigation of parameters in simulation-based dynamic traffic assignment models for traffic impact studies.
Loading...
Date
2025-09
Authors
Chin, Wei Ai
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Travel forecasting models are essential for anticipating traffic pattern changes and planning road infrastructure improvements. This study investigates the application of simulation-based Dynamic Traffic Assignment (DTA) using AIMSUN for traffic impact assessments (TIA), focusing on route choice calibration, behavioural parameter effects, and comparative model performance. Four route choice models namely Binomial, Proportional, Multinomial Logit, and C-Logit were calibrated and evaluated using the Geoffrey E. Havers (GEH) index based on simulated versus observed volumes. The Binomial model yielded the lowest GEH but lacked consideration of travel cost, limiting its use for impact assessment. The Proportional and Multinomial Logit models performed well when calibrated with K-Shortest Paths (K-SP) = 1 and optimal alpha or scale values. The C-Logit model was highly sensitive to scale and beta, with optimal performance at scale = 25 and beta = 0.1–0.15. The study also examined how driver cooperation, aggressiveness, and motorcycle composition influence delay in merging and signalised sections. In merging sections, regression analysis found both cooperation (β = –0.070, p < 0.001) and aggressiveness (β = –0.063, p = 0.019) to be significant predictors of delay. Cluster analysis revealed that lower cooperation levels (47.80%) were linked to significantly higher delays (1.81s), while decision tree analysis identified cooperation as the most influential factor, with a threshold at 45%, above which delays were reduced. In signalised sections, only motorcycle composition significantly influenced delay (β = 0.041, p < 0.001), where delay decreased from 20.25s to 16.78s as motorcycle share
increased. Decision tree results confirmed a threshold at 5.3% motorcycle share, above which delay rose sharply. The study compared traffic impact outputs from the DTA-based AIMSUN model and the static SIDRA model. The findings revealed that AIMSUN, particularly through its Time Series simulation, provided a more realistic representation of traffic conditions by capturing time-dependent variations, adaptive driver behavior, and localized congestion patterns. SIDRA, limited by static assumptions, consistently overestimated delays, especially at high-demand intersections. AIMSUN also better identified flow breakdowns and queue spillbacks, highlighting the strengths of DTA models for accurate and responsive assessments in complex urban networks. This study highlights the importance of route choice calibration, behavioural integration, and DTA application in improving TIA accuracy. The findings support the use of simulation-based methods in Malaysia’s urbanising areas and provide valueable insights that enhance the credibility of DTA models for more accurate forecasting and infrastructure planning for researchers, transportation planners, and engineers.