Connected Vehicle Data Enables Early Detection of Sydney Crash Hotspots

Recent research in Australia has applied connected-vehicle data and advanced statistical methods to identify locations in Sydney with a high risk of traffic incidents before crashes occur. These studies focus on using near-miss events and risky driving telemetry to support proactive road safety efforts.

One study examined spatial patterns of near-miss events and historical accident blackspots in Sydney using connected-vehicle data. Researchers applied spatial statistical techniques to determine where clusters of near-misses overlapped with areas of frequent crashes, providing a method to classify locations based on the joint occurrence of risky events and past accidents.

Building on this work, a subsequent study analyzed telemetry from more than 700,000 vehicles across New South Wales, collected between April 2020 and December 2021. The data included measurements of hard braking, harsh cornering, and rapid acceleration, with the most extreme 1% of events excluded to maintain consistency in the analysis.

This research categorized areas into four types based on crash and near-miss counts: High-High, High-Low, Low-High, and Low-Low. The studies consistently identified Sydney’s central business district, Parramatta, and Bankstown as locations with persistent high-risk driving behaviors.

What the numbers show

• Telemetry from over 700,000 vehicles was analyzed for risky driving patterns.
• Hard braking was defined as events exceeding 0.6 g, harsh cornering above 0.47 g, and harsh acceleration above 0.5 g.
• ARIMA achieved a mean absolute error of 162.21 in forecasting near-miss counts, closely followed by LSTM at 163.92.

The predictive study compared eight modeling approaches, including ensemble methods, deep learning, and classical time-series techniques. Results showed that the ARIMA model had the lowest mean absolute error, with performance similar to the LSTM deep learning model and better than the ensemble methods tested.

Spatio-temporal heatmaps generated from the data highlighted recurring risky driving hotspots in Sydney’s inner and western local government areas. These visualizations supported the identification of areas where proactive safety interventions could be prioritized.

Earlier research established a statistically significant link between near-miss clusters and official crash blackspots in Sydney. This finding validated the use of g-force telemetry as an indicator of locations where crashes are more likely to occur in the future.

The studies concluded that connected-vehicle telemetry can be used to support proactive road safety strategies. By identifying risky locations before crashes happen, authorities may be able to implement targeted interventions to reduce incident rates.

* This article is based on publicly available information at the time of writing.