Prevalence of high-risk behaviors among commercial motor vehicle drivers measured using artificial intelligence for naturalistic data collection.

Objective: Crashes involving commercial motor vehicles (CMV) result in high rates of injury and fatality, and rates have been increasing, garnering attention as a priority among transportation safety professionals. Major contributors to crash risk and fatalities, such as operator phone use, seatbelt noncompliance, and speeding remain insufficiently understood in terms of their prevalence, hindering the effectiveness of public health outreach and educational initiatives. This study used high-resolution cameras and Artificial Intelligence (AI) processing for naturalistic data collection to measure the prevalence of speeding, seatbelt noncompliance, and handheld phone use among commercial motor vehicle drivers in San Diego County, California. Technologic approaches utilizing AI can greatly expand understanding of the prevalence of these behaviors, allowing for improved opportunities to address this growing problem facing CMV operators, public health, and traffic safety professionals.

Methods: Using AI technology, radar and infrared cameras mounted on roadside trailers, data were collected continuously over 7-day periods at 16 locations across the county. The prevalence of CMV drivers' cell phone use, seatbelt noncompliance, and speeding was measured. More than 2,600 h of CMV driving data were collected anonymously across rural and urban locations, including on interstate and county roads, near the US/Mexican border, and on a Native American Reservation. Speeding was defined as exceeding posted speed limits of 55 mph on highways with a maximum speed of 65 mph; we examined both 55 mph (the general CMV speed limit in CA) and 65 mph as cutoffs for speeding. All cell phone and seatbelt violations identified by AI were manually reviewed for accuracy. Temporal associations by time of day, day of the week, and season, as well as roadway characteristics, were used to evaluate the propensity for these behaviors.

Results: Data were collected for 160,671 CMVs between April and August 2024. Of these, 17,341 (10.8%) demonstrated at least one risky driving behavior of speeding (65 mph cutoff), cell phone use, or seatbelt noncompliance. The most common risky behavior was speeding, 4.9% (n = 7195), followed by seatbelt noncompliance 4.5% (n = 7,143), and handheld phone use 2.6% (n = 4,241). The prevalence of all three offenses was highest between 6:30 AM and 8:30 AM (rush hour) and on weekends. The prevalence of speeding was 56.4% (n = 90,652) with a cutoff speed limit of 55 mph-the CMV speed limit in California.

Conclusions: Technological approaches can inform public understanding of the prevalence of behaviors that contribute to safety-critical mistakes. Offense prevalence was found highest on the lowest vehicular traffic days and times. These naturalistic data can guide safe driving policy, planning and decision-making as well as evaluate the impact of interventions while adhering to state and local privacy laws.