Abstract:

"A key component of Vision Zero is identifying and understanding the user attributes and needs that contribute to the incidence of deaths and serious injury in road events and that must, therefore, be taken account of in solutions. Understanding is dependent on observation and measurement to support scientific analysis, so it is inevitable that results should stagnate as existing observational methods reach their limits.

Traditional methods to identify potential crash therefore have a limit in terms of their ability to produce results and we have to start looking toward new techniques and applications. The use of probe in this application affords us the luxury of studying traffic pattern in detail with a high degree of accuracy and pattern matching. Instrumental to this research is EROAD, a technology company specialised in regulatory telematics, providing services in New Zealand, Australia and the United States. Through this partnership we were able to study over 10 million braking events across New Zealand and compare them to over 600,000 crashes ranging from minor injuries to fatalities.

Researchers implemented a number of geospatial analysis techniques such as DBSCAN, k-nearest neighbor, and voronoi calculations to cluster both harsh braking locations and crash locations and create a topographical model to assess similarity across time. The resulting clusters were further classified using a machine learning model combining publicly sourced data sets and data derived from EROAD relating to driver behaviour, fatigue, traffic volume and travel speeds. When applied to the New Zealand road network, the result is a risk model that highlights areas based on the likelihood of a crash occuring.
"