According to the US NHTSA, there were 38,680 fatalities in the US in motor vehicle crashes, in 2020 – up 7.2% from 2019. Connected Vehicle technology has emerged as a method to enhance motorists' safety and mobility. There are several Vehicle-to-Vehicle (V2V) technologies identified by the US NHTSA as candidates to improve roadway safety. To be effective, these technologies require a certain level of market penetration. Even with market penetration, there is a significant need for hazard prediction and prevention models to ensure the safety and mobility of motorists.
Researchers at the UCR Center for Environmental Research and Technology (CE-CERT) have developed a patented, agent-based, lane level hazard prediction application called Lane Hazard Prediction (LHP) that is based on available vehicle trajectories data collected from the V2V environment. The application identifies the downstream lane level hazard using spatial and temporal machine learning and data mining technique. provides an optimal lane decision recommendation so that the driver does not get into the lane with the hazard.
Framework for Lane Hazard Prediction application
The LHP application contains four major modules:
Mobile crowdsourced sensing – which obtains the position, speed, and direction of connected vehicles downstream.
Feature extraction – identifies the key factors that are critical for identifying a hazard or an abnormality in the traffic.
Lane hazard pattern recognition – provides a prediction on a lane level hazard position and its longitudinal bound.
Lane recommendation – provides recommendations on the optimal lane choice to avoid joining the traffic queued behind the hazard.
Name: Venkata Krishnamurty