Analysis and modeling of human driving behavior in highway merging areas using a combined approach of stochastic methods and artificial intelligence
ZHANG, XINZHE and Varga, István and Tettamanti, Tamás (2026) Analysis and modeling of human driving behavior in highway merging areas using a combined approach of stochastic methods and artificial intelligence. TRANSPORTATION RESEARCH INTERDISCIPLINARY PERSPECTIVES, 39. ISSN 2590-1982 10.1016/j.trip.2026.102213
![[img]](https://eprints.sztaki.hu/style/images/fileicons/text.png)
|
Text
Zhang_1_37490857_ny.pdf Download (4MB) |
Abstract
The paper investigates the driving behavior patterns of mainline human-driven vehicles during highway ramp-merging interactions. Using the exiD dataset, a dynamic merging window was defined in time and space to identify mainline vehicles with potential interactions with ramp vehicles. Using the neighboring-vehicle information for each ramp vehicle within the merging window, one-to-one merging events between mainline and ramp vehicles were extracted. To identify heterogeneous driving behavior patterns within the merging area, a Bayesian Gaussian Mixture Model was employed for cluster analysis of driving behaviors. Subsequently, the Hierarchical Bayesian Inverse Reinforcement Learning framework was adopted to infer global and cluster-specific bias weights that represent relative reward preferences across behavioral patterns. Afterward, using reinforcement learning, the driving behavior trajectories of human-driven vehicles in each cluster were reproduced in SUMO (Simulation of Urban MObility), providing a more comprehensive representation of lateral and longitudinal driving behavior. Finally, the exiD-trained policy was deployed in a high-fidelity SUMO merging scenario representing the Hungarian M1 highway. The results showed stable closed-loop execution and supported the cross-scenario portability of the framework. Overall, this study presents a configurable empirical framework for on-ramp merging analysis that connects traffic interaction representation, human driving behavior heterogeneity, data-driven preference inference, and closed-loop traffic simulation.
| Item Type: | Article |
|---|---|
| Subjects: | Q Science > QA Mathematics and Computer Science > QA75 Electronic computers. Computer science / számítástechnika, számítógéptudomány |
| Divisions: | Systems and Control Lab |
| SWORD Depositor: | MTMT Injector |
| Depositing User: | MTMT Injector |
| Date Deposited: | 29 Sep 2026 11:11 |
| Last Modified: | 29 Sep 2026 11:11 |
| URI: | https://eprints.sztaki.hu/id/eprint/11166 |
 |
Update Item |
