Ashutosh Arun is this year's winner!
Congratulations to Ashutosh Arun of the School of Civil and Environmental Engineering, Queensland University of Technology for winning the 2021 SIDRA SOLUTIONS Postgraduate Award for his PhD thesis titled ‘A Novel Road User Safety Field Theory for Traffic Safety Assessment Applying Artificial Intelligence-Based Video Analytics’.
Summary of Reseach
Ashutosh's research focuses on non-crash-based safety assessments of transport facilities through surrogate measures of safety. Conventional safety assessment techniques are reactive and rely heavily on police-reported road crash data that has several shortcomings such as under-reporting, low sample means, limited behavioural information, and omitted variable bias. Also, waiting for crashes to accrue at a location BEFORE any solutions could be implemented is unethical. This research makes use of the latest advances in video analytics and proposes a new Road User Safety Field (RUSF) Theory to analyse safety-relevant road user behaviour and proactively estimate both crash frequency and severity at a transport facility. The Road User Safety Field theory argues that crash risk is a fundamental property of road user movements and borrows from the physics concept of energy fields like the electromagnetic field to mathematically define the safety “buffer” that road users typically maintain around them while moving in a road environment. This safety field essentially represents the risk appetite or the inherent risk proneness of a road user. When two road users interact, their safety fields overlap, resulting in a psychological repulsive risk force that helps them maintain safe distances. The risk force can also be harnessed to model the crash risk involved in any traffic interaction, making it a potent surrogate measure for non-crash-based assessments of transport facilities such as signalised intersections.
This research first introduced novel methods to estimate crash frequency by severity levels using common traffic conflict measures such as Time-to-Collision and Delta-V observed from four signalised intersections in Southeast Queensland. An automated process was used to extract traffic conflict measures from traffic monitoring videos using computer vision models. Finally, the RUSF-based risk force models were integrated into the joint crash frequency and severity modelling framework and compared with other prevalent traffic conflict measures such as Modified Time-to-Collision and Deceleration Rate to Avoid a Crash. The analysis found that the RUSF model estimated the crash risk more accurately and precisely in real-world traffic scenarios than the current methods. The study results significantly improve the effectiveness of automated safety analysis for transport facilities and could transform long-term road designs as well as the safety prediction algorithms of real-time infrastructural applications like adaptive signal control systems and Connected and Automated Vehicles (CAVs).