Samson on his research

Modern traffic models are often complex and implemented with many lines of computer codes. Their predictions often crucially depend on many parameters. Estimating the appropriate parameter values from data, also called model calibration and validation, is a difficult inverse problem with many subtleties.

Samson’s project aims to develop statistically rigorous methodologies to improve traffic model calibration practices. In particular a Bayesian approach was adopted, which is flexible and incorporates proper uncertainty quantification.

What questions will this research answer?

Unlike machine learning models that typically focus on predictive accuracy, traffic models are supposedly interpretable. Thus, fitting traffic models to data and drawing reliable inferences requires more statistical endeavour.

The goal of Samson’s work is to develop advanced mathematical and statistical methods for estimating traffic model parameters, and to demonstrate their applicability on selected traffic models.

Why is this work important?

It provides guidance on traffic modelling practices, which can influence investment and design decisions. It is hoped that the results, if applicable, can be incorporated into official traffic modelling guidelines.

As such it will help improve traffic model performance so more-informed decisions can be made.

What drew Samson to this work?

Throughout his undergraduate studies, with a double major in civil engineering and mathematics, Samson was fascinated with the creativity in mathematics and the real-world applicability of engineering.

This transport modelling project is a unique and interesting opportunity for me to combine the skills that he has acquired to date.

How far along is Samson in his PhD?

At the time of writing Samson is in the fourth year of his PhD, busy with writing his thesis, which he hopes to submit soon.

He has successfully implemented computer programs to run state-of-the-art Bayesian computation algorithms, and also developed simulations for popular traffic models such as car following and gap acceptance.

In terms of external audiences, he has published one journal article, and has presented his research at several conferences.

What are the big lessons of his work so far?

A key challenge that Samson faced was that his initial methodology was inconclusive and lacked statistical rigour. He did not have the mathematical and statistical background for the project.

Fortunately, his supervisors encouraged him to attend a winter school, from which he had the opportunity to learn about what would become the key statistical skills needed precisely for the project.

Since then, he has spent a lot of time learning about the state-of-the-art statistical modelling techniques and was able to apply them to the project successfully.

What personal / professional skills has he developed?

One of the professional skills that Samson’s PhD has helped him develop is effective communication for audiences of different backgrounds. He has had various opportunities to present his research to different audience including industry representatives, researchers with postgraduate mathematical background but not specialising in statistics, and researchers with a statistics-focused background.

The result of which is he has learnt about tailoring the content and style of his presentations for more effective dissemination of research aim and outcome.

What does Samson see as his next move upon completing his PhD?

Right now, Samson is uncertain whether he’ll stay in academia or enter the industry, as both have their pros and cons. In any case, he feels his PhD has been an incredible journey from which he has picked up a lot of transferrable skills.

He plans to continue to develop his mathematical and statistical skills, and to apply them to solve important real-world problems.

A word from PhD supervisor, Dr Chao Sun

Why is this PhD important to investigate?

Historically, model calibration has been an under-researched area, often due to a lack of rich observational data. Now, we have a significant amount of data, but understanding how to effectively use it for model improvement isn’t always clear-cut.

Samson’s work aims to bridge this gap, helping us make more accurate predictions and, as a result, design more efficient roads and intersections.

What are the major challenges to overcome in the field?

Absolutely! For instance, our video analytics data was still under development when Samson began his PhD. His meticulous error checking and insights aided us in refining the results.

Additionally, the integration with third-party modelling packages poses challenges. Samson’s impressive programming skills have allowed him to bridge those gaps, developing tools to seamlessly integrate them into his research.

His adaptability and relentless drive to acquire the necessary skills underscore why he is so well-suited for this work.

Where might this work lead in the (near and far) future?

Our hope is that the insights from Samson’s work can be applied more broadly to improve transport/traffic models.

Contact Samson

If you’d like to contact Samson about his research, contact can be made via his LinkedIn profile.

Publications

Parameter estimation for Gipps’ car following model in a Bayesian framework – Physica A: Statistical Mechanics and its Applications. Volume 639, 1 April 2024, 129671

Conference presentation

Parameters Estimation for Car Following Model Using Bayesian Inference, at 10th International Congress on Industrial and Applied Mathematics, ICIAM 2023.