ESR13 – Computational Vehicle/Pedestrian Interaction Models

I am Amir Hossein Kalantari, ESR13 and PhD student at ITS, University of Leeds. I received my BSc degree in civil engineering and M.Sc. degree in transport engineering in 2015 and 2018, respectively. I started taking interest in human factors and traffic psychology in 2015 when I started to investigate the prevalence of distracted driving, the related psychosocial factors, and its association with drivers’ at-fault accidents among Iranian drivers employing mathematical models. Since 2020, I have been working on modelling and simulating road user interactions using both naturalsitic and lab data.

I am Amir Hossein Kalantari, ESR13 and PhD student at ITS, University of Leeds. I received my BSc degree in civil engineering and MSc degree in transport engineering in 2015 and 2018, respectively. I started taking interest in human factors and traffic psychology in 2016 when I opted to investigate the prevalence of distracted driving, the related psychosocial factors, and its association with Iranian drivers’ at-fault accidents employing mathematical models. Since 2020, I have been working on modelling and simulating road user interactions using both naturalistic and lab data.

My affiliation

Contact details to me and my supervisors:

University Profile Page | ResearchGate | LinkedIn | Scholar

Prof Gustav Markkula: G.Markkula@leeds.ac.uk

Prof Natasha Merat: N.Merat@its.leeds.ac.uk

Prof Marco Dozza: Marco.Dozza@chalmers.se

Background

Pedestrians constitute a great proportion of the traffic ecosystem and their interaction, as vulnerable road users (VRUs), with others such as drivers (vehicles), has a great impact on traffic safety and efficiency [1]. With recent advancements in vehicle automation, the challenge of human-robot interaction has arisen, making it necessary for automated vehicles (AVs) to communicate their intentions and negotiate driving strategies with pedestrians, such as right of way [2]. Effective communication between AVs and pedestrians, as well as the ability of AVs to respond to changes in pedestrian behaviour, can not only increase trust in them and their driving performance but also minimise their potential conflicts at unsignalised locations [3]. Thus, it is crucial to understand both competing and communication strategies that exist between pedestrians and drivers/vehicles to achieve a safe, efficient and transparent traffic flow. To this end, previous research has strived to investigate vehicle-pedestrian interactions employing mathematical models such as game theory but it is still unclear whether conventional game theory is enough for modelling such interactions or whether more complex paradigms such as behavioural game theory are essential for this.

Aims and objectives

The main objective of this PhD project is to employ game-theoretic (GT) models to see how pedestrians interact with vehicles (or AVs) in different crossing scenarios. In order to achieve this, the following sub-objectives are defined:

  1. Identifying the proper modelling candidates to propose a computational framework of road user interaction.
  2. Planning, designing and conducting controlled studies using human-in-the-loop simulated environments to provide validation tools for game-theoretic models.
  3. Planning and conducting naturalistic studies.
  4. Comparing the findings of the controlled study with the naturalistic data and using both datasets to improve the computational framework performance.

Research description

  1. Several GT models were selected from the literature and their suitability for further revision/extension was considered. Finally, a conventional [4] and a behavioural GT model [5] were chosen for the computational framework. The conventional GT model was tested for its convenience in generating pedestrians’ gap acceptance behaviour using a distributed simulator study (DSSd) dataset where a desktop driving simulator was connected to a CAVE-based pedestrian lab [6].  Overall, five computational models were considered for the framework: 1) The conventional GT model was solved by the mixed-strategy Nash equilibrium. 2) The payoff formulation of the conventional GT model was revised to correct some of the assumptions of the model that it was believed to hinder the model from fully capturing road user behaviour. The model then was solved by the mixed-strategy Nash equilibrium. 3) The payoff formulations of the models in (1) and (2) were also solved by a behavioural GT model known as the dual accumulator model [5] creating two more model variants. 4) Finally, a logit model was tested and considered in the model comparisons [7].  
  2. A DSS was designed and conducted by connecting a high-fidelity motion-based driving simulator to a CAVE-based pedestrian lab. In this study, an experimental paradigm was proposed in which two human agents could interact with each other in a safe and controlled environment. Thirty-two pairs of one driver and one pedestrian interacted with each other under different traffic scenarios and the effect of their demographic characteristics, personality traits as well as road infrastructure’s priority rules and vehicle kinematics on a number of interaction-related metrics (e.g. pedestrians’ choice to pass first) was studied [8]. You can watch the video of the study below:

3. Two marked crossings in city of Leeds were surveyed to collect real-time traffic data. Following several roadside observations and consultations with Leeds City Council, the locations were chosen based on safety concern and the prevalence of one-to-one interactions between vehicles and pedestrians. Two Viscando camera sensors were used to collect data over 14 days (seven days for each location). The sensors detect road users (light vehicles, heavy vehicles, cyclists, pedestrians) and track their trajectory and speed over discrete time stamps.

Results

  • DSSd could generate a gap acceptance dataset with respect to both AV and HD (human-driven) conditions that is close to reality [6].
  • DSS could simulate scenarios where traffic agents interactively communicate with each other, demonstrating behaviours that are qualitatively in line with those observed in naturalistic studies. The findings showed that kinematic cues, including vehicle speed and time gap, had a stronger influence on pedestrians’ crossing behaviours at unmarked crossings, than personality traits such as sensation seeking and social value orientation [8].
  • The proposed behavioural GT model (the revised payoff solved by behavioural GT) outperformed other models in predicting interaction outcomes as a function of kinematics and crossing location type. The model, unlike conventional GT and logit models, is capable of predicting which interaction will take the longest to resolve. Overall, the findings suggest that road users cannot be expected to play the Nash equilibrium of conventional GT [7].

My Publications

Journal

Kalantari, A. H., Yang, Y., de Pedro, J. G., Lee, Y. M., Horrobin, A., Solernou, A., … & Markkula, G. (2023). Who goes first? A distributed simulator study of vehicle–pedestrian interaction. Accident Analysis & Prevention, 186, 107050. https://doi.org/10.1016/j.aap.2023.107050

Conferences/Posters

Kalantari, A. H., Yang, Y., de Pedro, J. G., Lee, Y. M., Horrobin, A., Solernou, A., … & Markkula, G. (August, 2023). A distributed simulator study of car-pedestrian interaction. In 7th International Conference on Traffic and Transport Psychology, Gothenburg, Sweden.
Kalantari, A. H., Markkula, G., Uzondu, C., Lyu, W., Garcia de Pedro, J., Madigan, R., … & Merat, N. (2022).  Vehicle-Pedestrian Interactions at Uncontrolled Locations: Leveraging Distributed Simulation to Support Game-Theoretic Modeling. Transportation Research Board (TRB) 101st Annual Meeting. Washington D.C., USA.  (No. TRBAM-22-01874).

Kalantari, A. H. (2021). Game-theoretic modelling for vehicle-pedestrian interactions. Poster presented at the 13th International Conference on Automotive User Interfaces and Interactive Vehicular Applications. https://www.shape-it.eu/wp-content/uploads/2021/09/ESR13_Amir-Hossein-Kalantari_v2-2.pdf

Prepritns

Zhang, C., Kalantari, A.H., Yang, Y., Ni, Z., Markkula, G., Merat, N. (2023). Cross or Wait? Predicting Pedestrian Interaction Outcomes at Unsignalized Crossings.  https://doi.org/10.48550/arXiv.2304.08260
Kalantari, A.H., Yang, Y., Merat, N., & Markkula, G. (2023, March 17). Modelling vehicle-pedestrian interactions at unsignalised locations: Road users may not play the Nash equilibrium. https://doi.org/10.31234/osf.io/axseu

Markkula, G., Lin, Y., yslin, Srinivasan, A. R., Billington, J., Leonetti, M., Kalantari, A. H., … Merat, N. (2022, June 8). Explaining human interactions on the road requires large-scale integration of psychological theory. https://doi.org/10.31234/osf.io/hdxbs

References and links

[1] W. H. Organization, ‘Pedestrian safety: a road safety manual for decision-makers and practitioners’, 2013. https://apps.who.int/iris/rest/bitstreams/279316/retrieve

[2]  P. Koopman and M. Wagner, ‘Toward a framework for highly automated vehicle safety validation’, SAE Tech. Pap. Tech Rep, 2018.

[3]  J. E. Domeyer, J. D. Lee, H. Toyoda, B. Mehler, and B. Reimer, ‘Interdependence in vehicle-pedestrian encounters and its implications for vehicle automation’, IEEE Trans. Intell. Transp. Syst., vol. 23, no. 5, pp. 4122–4134, 2020.

[4] W. Wu, R. Chen, H. Jia, Y. Li, and Z. Liang, ‘Game theory modeling for vehicle–pedestrian interactions and simulation based on cellular automata’, Int. J. Mod. Phys. C, vol. 30, no. 04, p. 1950025, 2019.

[5] R. Golman, S. Bhatia, and P. B. Kane, ‘The dual accumulator model of strategic deliberation and decision making.’, Psychol. Rev., vol. 127, no. 4, p. 477, 2020.

[6]  A. H. Kalantari et al., ‘Vehicle-Pedestrian Interactions at Uncontrolled Locations: Leveraging Distributed Simulation to Support Game Theoretic Modeling’, 2022. (No. TRBAM-22-01874). https://trid.trb.org/view/1909546

[7]  A. H. Kalantari, Y. Yang, N. Merat, and G. Markkula, ‘Modelling vehicle-pedestrian interactions at unsignalised locations: Road users may not play the Nash equilibrium’, 2023. [Online]. Available: psyarxiv.com/axseu.

[8] A. H. Kalantari et al., ‘Who goes first? a distributed simulator study of vehicle–pedestrian interaction’, Accid. Anal. Prev., vol. 186, p. 107050, 2023, doi: https://doi.org/10.1016/j.aap.2023.107050.

The following video reflects the initial view of the project: