Calibration and validation of a hybrid traffic flow model based on vehicle trajectory data from a field car-following experiment

ABSTRACT

The paper focuses on the calibration of the hybrid multi-scale modelling approach that incorporates multiple levels of traffic flow representation. This paper refers to a model already developed in the literature, the Hybrid Cellular Automata (CA) and Cell Transmission Model (CTM). We use vehicle trajectory data collected from a car-following field experiment on a circular road track to explore the calibration of the CA model concerning various cell lengths through two distinct approaches: simulating all vehicles within the closed loop and simulating each vehicle using data obtained from its respective follower. We also evaluate different methods for the CTM calibration with respect to the CA model. The major findings are: (1) the calibrated parameters obtained using the simulated leader approach display greater regularity across different cell lengths; (2) the Constrained Squared Error [CsqE] method for macroscopic calibrations yielded promising results, showcasing the lowest sum of squared errors between fundamental diagrams.

KEYWORDS:

• Cell Transmission Model
• Cellular Automata
• fundamental diagram
• experimental data

Acknowledgements

This manuscript reflects only the authors’ views and opinions, neither the European Union nor the European Commission can be considered responsible for them. Authors wish to thank anonymous reviewers for their helpful comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 That is, the traffic state that the driver will experience (Cascetta 2009; Cantarella et al. 2019).

2 The proposed model was preliminary calibrated using the data collected on 400 meters of the 5 I80 highway in the USA (NGSIM I80-1, 2021), during 15 min. This section of the highway has six lanes, named from 1 to 6 from left (fastest) to right (slowest), and an incoming merging ramp, called lane 7.

3 The repeated simulations approach has been discussed in several studies (Laval, Toth, and Zhou 2014; Tian et al. 2016); as suggested by Tian et al. (2021). In this paper, initially 200 simulations were considered. However, after a preliminary calibration of the model, it was observed that the results stabilized after 100 simulations. Therefore, this number was adopted as the final setting.

4 For this analysis, the number of simulations has been increased, since the parameters are fixed and the second indicator considered depends on the single best trajectory found for each vehicle, instead of the aggregated result of the mean RMSE value used during the calibration process.

Additional information

Funding

This study was carried out within the MOST – Sustainable Mobility National Research Centre and received funding from the European Union Next-GenerationEU (PIANO NAZIONALE DI RIPRESA E RESILIENZA (PNRR) – MISSIONE 4 COMPONENTE 2, INVESTIMENTO 1.4 – D.D. 1033 17/06/2022, CN00000023). This research was partially funded by the University of Salerno, under local grants no. ORSA214124-2021, no. ORSA223793– 2022 and no. ORSA238719 –2023. It was also partially funded by the project DIGIT-CCAM ‘Progetti di Rilevante Interesse Nazionale' (PRIN 2020 – MUR), and I CAN BE ‘Progetti di Rilevante Interesse Nazionale' (PRIN 2022 – MUR).