The future of road safety is uncertain. Despite the push toward increasing road safety in Europe, the fact remains that road crashes are one of the leading causes of death. Road safety campaigns often centre on reducing aggressive driving behaviours, reducing driving while cognitively impaired, and ensuring the vehicle’s safety-critical functionality is well maintained. These campaigns are often set in motion as a means to meet ambitious targets for collision rate reductions. Although a level of risk reduction is often achieved, the ambitious targets are often missed. Giving us even greater hope of a safer future is the increasing prominence of advanced driver assistance systems (ADAS) and automated vehicles on the roads, which are expected to result in an appreciable reduction in collision rates. Yet, despite these advancement, road crashes will remain a highly random and non-deterministic process.

 

With that in mind, this research stream introduces a forecasting tool to embrace the non-determinism of this uncertainty, and provide reasonable predictions for setting and evaluating safety targets. The tool that is introduced is an extended version of the Heston Stochastic Volatility model; a common modelling technique employed in mathematical finance. It is favoured as it leverages the evolution of two interconnected yet randomly varying processes to investigate how an asset price may change over time. Similar assumptions can be applied to road safety: road crash numbers are intrinsically linked to the volume of vehicles on the road.

The first study along this stream was introduced to be used as an additional tool by policy analysts to enhance the accuracy of short-term predictions on collision rates, and analyse long-term safety target proposals. The second study along this stream was introduced to demonstrate the model’s ability to adapt to local nuances, in regions where crash rates are frequently affected by sporadic deviations from the norm.

I was recently featured on Data Skeptic talking about my research in this area.

RELEVANT WORKS

• Shannon, D. and Fountas, G. (2022) 'Amending the heston stochastic volatility model to forecast local motor vehicle crash rates: A case study of Washington, DC', Transportation Research Interdisciplinary Perspectives, 13, 100576. DOI: 10.1016/j.trip.2022.100576 

• Shannon, D. and Fountas, G. (2021) 'Extending the Heston model to forecast motor vehicle collision rates', Accident Analysis & Prevention, 159, 106250. DOI: 10.1016/j.aap.2021.106250