Pairing ITS Infrastructure with CV Data for Insights into Traffic Safety Events
Traffic sensors are a valuable tool for traffic management agencies to obtain accurate data related to vehicle counts, occupancy, and even vehicle classification. However, this technology does not come without its drawbacks. Sensor hardware as a whole involves costly installation and maintenance, often requiring a pavement cut or mounted pole to be erected. On average, traffic sensors can range anywhere from $500 – $26,000 on initial purchase, with the addition of regular maintenance and repairs. Furthermore, certain sensor types perform poorly when subjected to extreme weather such as fluctuating temperatures, strong winds, dense fog, and heavy rain.
Though not a replacement to roadway sensors, connected vehicles (CVs) are a valuable, cost-effective supplement. In the next two years, estimates show that 90% of new vehicles in the US will be shipped with embedded connectivity, meaning they can share anonymous data on speeds and volume. CVs generate large amounts of data; one estimate states that each CV generates a massive 280 petabytes of data annually. There are diamonds of information in this data. However, like diamonds mined from tons of ore, the key insights need to be found, extracted, analyzed and translated before they can become actionable information.
This case study argues that leveraging data from connected vehicles is a worthy investment for traffic management agencies. Without physical installation or regular maintenance, CVs can provide a world of insights into problematic roadways, traffic patterns, driver speeds, and more.
The following sections concentrate on a specific use-case for leveraging CV technology: identifying sudden changes in a driver’s velocity such as harsh braking and acceleration. Waycare labels these abrupt variations as safety events.
Safety Events
Harsh braking occurs when a driver uses more force than usual to slow down their vehicle, often in an attempt to avoid a collision. Conversely, harsh acceleration is characterized by a sudden increase in speed. The presence of these sudden velocity changes can indicate a problem on a roadway, such as an area of irregular congestion or an obstacle/debris.
These variable roadway conditions are often a major blindspot to traffic management agencies, since they are unpredictable and traditional infrastructure has not been designed to detect their presence. CV data can be used as a supplement to sensor data to help traffic management agencies identify and further investigate problematic road conditions.
Method
Waycare receives anonymized CV data from its various partners in the city of Las Vegas. Data Scientists at Waycare looked for large speed differentials in the data set and defined certain thresholds for their analysis. Patterns of harsh braking and acceleration from July 2019 through June 2020 are detailed below.
Findings
Figure 1 shows the number of safety events over the course of a year during which data collection took place. These numbers were adjusted based on the variation of incoming monthly vehicle data. There is a notable increase in safety events starting in September compared to the two months prior. This bump can possibly be attributed to increases in traffic congestion as a result of students going back to school. There is also a very noticeable drop in safety events in the months following February. The significant decline in traffic congestion and crashes due to statewide shutdown orders for the COVID-19 pandemic are the cause.
Figure 1: Safety Events Per Month
Numbers were normalized per 10,000 vehicles due to the changing number of vehicles per month.
Major acceleration event – increase in speed of 7.44mph/second & above
Major deceleration event – decrease in speed of 11.47mph/second & above
The heat map in Figure 2 shows areas of harsh braking and crashes from July – October 2019. This provides a visualization of the volumes of harsh braking and traffic crashes in different locations along I-15, US-95, and I-215. Road segments with more significant clustering indicate that there are more abrupt driving activities, whereas segments with fewer points indicate potentially safer traffic conditions. These segments can be broken down into more granular views for closer inspection.
Figure 2: Harsh Braking vs. Traffic Crashes
Time Period: July – October, 2019
In Figure 3, we have identified a positive correlation between locations where harsh braking and crashes have occurred. While correlation does not prove causality, we can surmise that areas with a greater presence of harsh braking potentially indicate dangerous conditions, that may lead to a greater likelihood of traffic crashes.
Figure 3: Harsh Braking and Traffic Crash Correlation
Conclusion
The examples provided in this case study demonstrate the ability of CV data to shine a light on potentially dangerous or problematic roadways. Abrupt driving behavior is a symptom of a larger problem. It can alert traffic agencies of dangerous road conditions like irregular traffic congestion or road obstacles, and serves as a resource for subsequent investigation. Furthermore, CV data can be leveraged as a virtual network, without requiring the maintenance or upgrades associated with traditional infrastructure.
Related Posts
January 19, 2021
Hiding in Plain Sight – How can AI help us identify non-recurring congestion?
Read ArticleDecember 18, 2018