Machine vision, AI to improve fleet driving

  • March 5, 2020
  • imc

Haulage and bus telematics specialist Lytx has launched technology to tackle distracted driving with machine vision and AI-powered risk detection. The new technology is aimed at improving safety in commercial, public sector and field services fleets. The new tech can provide fleet managers with near-real-time insight into such risky behaviours as texting while driving.

According to San Diego, California, based Lytx, its new technology acts like a co-pilot, constantly monitoring the road and the driver, detecting risky driving and triggering short video clips to provide insight into what’s really going on inside the vehicle. 

Lytx’s risk detection technology is included as part of its Driver Safety Suite and includes four new combined machine vision and AI (MV+AI) triggers: the driver holding a handheld device while driving, not using a seatbelt, consuming food or drink, and the driver smoking. 

It can be coupled with Lytx’s Fleet Tracking Service to provide what the company describes as “a complete and integrated solution for fleets looking to optimise their vehicle technology with a single vendor”.

“Because of the vast scope and scale of our driving data, we have advanced the use of MV+AI to unprecedented levels within the industry,” said Brandon Nixon, Lytx chairman and CEO. “We now have the ability to more fully and accurately capture and identify risk for the fleets we serve even better than before.”

Lytx claims to achieve greater than 95% accuracy detection across more than 60 risky driving behaviours. In 2019, Lytx labelled over 1.75 million minutes of video with cell phone use, driver unbelted, smoking and food/drink behaviours. 

The company says that it has analysed 120 billion miles of driving data, and captures 100,000 risky driving events daily. It has used this data as the basis for developing its machine vision and AI risk detection technologies, as well as using human analysts. The technology employs algorithms to identify the data points that are most important in building an accurate picture of risk, resulting in prescriptive and customised coaching insights to fleet managers.

The new triggers are an extension of Lytx’s existing MV+AI road-view triggers launched in 2015: rolling stop, lane departure, following distance and critical distance. Both sets of triggers monitor for driving behaviour patterns and select short video highlights to be analysed and screened by Lytx’s artificial intelligence for accuracy. Validated video clips are then automatically sent via a 4G LTE cellular connection from the vehicle to the cloud, where they may be viewed by a client via their Lytx account from any internet-connected computer, smartphone or tablet.

Fleet managers and drivers have the opportunity to view these video clips using Lytx’s proprietary coaching workflow, a program designed to help change driving behaviour. Lytx says that it is associated with up to 50% reduction in collisions and 80% reduction in associated claims costs.

“Texting while driving and driving without a seatbelt, as well as industry-specific use cases, like smoking, eating or drinking while driving, are critical risky behaviours for clients,” said Jim Brady, Lytx vice president of product management. “Other video telematics providers that attempt to detect these types of behaviours either bring back a large percentage of false positives or burden clients with mountains of data. By comparison, Lytx’s MV+AI algorithms are the most advanced in the industry, so our clients only receive the most important clips that represent key coachable moments in a format that doesn’t overwhelm them.”