Can we compare the driving habits in the U.K. (United Kingdom) and U.S. (United States) and get a valid picture of how to improve fleet operations and driver safety? Is there a way AI (artificial intelligence) and machine vision can help to eliminate accidents?

A vendor of machine vision and AI-based telematics, Lytx in England, has been collecting data on fleet operations in the U.K. and issued a report recently that analyzed risky driving trends among fleets. Lytx’s database contains more than 120 billion miles of driving data from more than one million commercial drivers worldwide, capturing over 64 million risky driving events globally each year.

Zeroing in on their homeland, Lytx found U.K. fleet drivers exhibited these dangerous behaviors: driving too close to the vehicle ahead, slow or late response, mobile phone use while driving, driving without a seatbelt, and failure to maintain proper space around the vehicle so that the driver has an escape route should the unexpected occur. In analyzing the data, Lytx found these behaviors are directly related to potential collisions, proving that the elimination of habitual risky driving behaviors can have an immediate and lasting effect on the frequency and severity of collisions and near collisions.

A driver who follows the vehicle in front of them too closely—this being the most prevalent risky driving behavior—is approximately 40% more likely to have a collision in the next 90 days than a driver who ensures proper following distance. Similarly, a driver who demonstrates late response (slow reaction) to a potential hazard is 80% more likely to have a collision within 90 days than a driver who responds within a proper time frame.

While not in Lytx’s top five, impaired driving has long been a behavior issue in the U.K. and U.S. Public concern about impaired driving in general and particularly drug-impaired driving has increased along with distracted driving. As a result, the need for the development of early warning systems for driver monitoring is getting attention for preventing collisions and fatalities.

With the increased number of states that have legalized cannabis, driving under the influence of marijuana is becoming more of an issue. The ACS (alcohol countermeasure systems) has research the development of non-intrusive artificial intelligence technology to screen for cannabis impairment. Based on research conducted in collaboration with the CAMB (Centre for Addiction and Mental Health), ACS has designed and developed a Cannabis Screening System using AI models and eye-tracking information.

The results show that several eye tracking features highly correlate with one or both target measurements of blood THC level and SDLP (standard deviation of lateral position) measurements. The potential of eye tracking information to be used as a non-intrusive marker of THC presence and/or impairment for driver monitoring is impressive.

The research results also show that both THC presence and instances of impaired driving can be detected with high accuracy (95% for THC and 81% for impairment detection) using deep learning methodologies. With approximately 24% of drivers reporting driving within two hours of cannabis consumption and with increased accessibility of cannabis through legalization, a reliable Cannabis Screening System can be used for in-vehicle driver monitoring and continuous real-time evaluation of driver vigilance to reduce impaired-driving incidents in fleets.

Want to tweet about this article? Use hashtags #construction #IoT #artificialintelligence #5G #cloud #machinelearning #bigdata #digitaltransformation #cybersecurity #blockchain #sustainability #future of work #infrastructure #driverdistraction

Click here to read more News Analysis.