By Fleet Management Weekly Staff
May 9, 2023
Collisions and crashes can have a devastating impact on a company’s bottom line. Higher premiums, property damage, injuries to third parties, and worker injury can cost fleets millions of dollars per year. While there are many safety programs on the market designed to lower driver risk, one company is approaching it in a new way: Artificial Intelligence (AI).
While today’s fleet safety programs are addressing driver safety through ‘connected vehicle’ technology that identifies risky driver behavior and data analyses, the ability to predict risky behavior is still in the formative state when it comes to situational awareness.
Nauto CEO Stefan Heck believes that the company is well-positioned to extend the driver safety envelope with a real-time, AI-enabled driver and fleet safety platform. “By analyzing billions of data points from over three billion AI-analyzed miles, our machine learning algorithms continuously improve and can predict and improve driver behavior before events happen,” he says.
Stefan backs up his assertion of Nauto’s platform’s effectiveness by noting that its AI-powered fleet management software’s ability to anticipate driver behavior has enabled large commercial fleets to avoid more than 30,000 collisions and saved nearly $300 million. He notes, “Through analysis of factors inside AND outside the vehicle, we can uniquely determine the likelihood of a collision and alert drivers of imminent risks to avoid it.”
The Impact of Collisions by Fleet Size
Collisions and crashes impact a company’s bottom line in a few ways. First, if the company is insured by a third-party insurer, rates can go up, plus a deductible. As fleets get larger, the amount of collision damage costs goes up.
“Most mid-size fleets of around 2,000 vehicles often have to pay the first 100K-to-1M dollars,” says Stefan. “Large fleets with over 3,000-4,000 vehicles are usually fully self-insured, so every collision avoided goes straight to the bottom line.
“In addition to property damage to the vehicle, the much larger portion of the bill is injuries to people (both the drivers themselves and third parties). If the driver is an employee of the fleet and not a contractor or temporary worker, then any injury they sustain is also a worker’s comp claim. When adding up all those elements, most fleets pay between $3,000-$10,000 of loss per vehicle per year. It’s usually the fleet’s second or third biggest cost (behind driver salaries)”
Collisions also have a direct operational impact beyond the actual damage and injuries caused, according to the Nauto CEO: “In a package delivery business, every time a delivery driver crashes, there’s a set of packages that are late to be delivered, if they even get delivered at all.”
There’s also a brand impact, particularly for bigger companies with famous brands. Collisions are also one of the top causes for drivers to quit their jobs. Many fleets are struggling to find enough drivers, and the more collisions you have, the more drivers will quit either due to medical problems or out of fear of future collisions.
Augmenting Data from Others Solutions
Most fleets that Nauto work with already have safety programs in use – usually telematics, training, or MVR monitoring programs, but Nauto expands the company’s overall safety purview, alerting fleets to variables that are much more difficult to track, like distraction.
“Many safety programs on the market have good results, especially the ones that focus on new drivers,” says Stefan. “But we still see high-risk rates for factors that people don’t track, such as distraction. Distraction usually doesn’t show up in your MVR and is massively under-reported. After all, drivers usually don’t self-report when they use their phone while driving. Until you have a computer vision AI system, you won’t know how distracted your drivers are.”
Because Nauto has visual detectors of collisions, companies can see gentler impacts like backing into a mailbox or hitting a pedestrian. Stefan notes that fleets often have three or even four times as many collisions as they thought they did: “We see really high rates when we go into a fleet, usually between four and six distractions per moving hour, although we’ve seen it as high as 26. We can generally reduce the distraction rate by at least 80% and sometimes can eliminate distractions altogether.”
Stefan notes that Nauto has seen bigger reductions with the AI-powered system than one would with the traditional methods of just training or coaching: “Compared to telematics, which will get you 10-15% reduction, or traditional cameras, which get you between 18-25%, we can generally achieve two-thirds reduction. In some fleets, we’ve gone as high as 90% reduction, but it depends on how risky your fleet was to begin with.”
Predictive AI: Anticipating a Crash
According to Stefan, predictive AI isn’t just observing what’s happening, but is actually anticipating what will happen next. This is very different from telematics in that it helps the driver anticipate what happens next instead of just examining what’s already happening.
Nauto uses predictive AI to examine what’s happening around a vehicle, such as if you’re approaching an intersection or the light is changing. It also looks at what’s happening inside the vehicle, such as if the driver is distracted or if their hands are on the wheel. Then the algorithm fuses the information in real-time to determine if the combination is risky.
“For example, if you’re looking at your phone at a red light, your risk of getting rear ended is modestly increased, but you’re not going to run into anything,” says Stefan. “If you’re looking at your phone while merging onto a freeway and then the lead vehicle breaks while you’re already tailgating, you’re talking about a thousand times the normal driving risk.
We use 30 different real-time preventive AI algorithms that run directly onboard the vehicle. They automatically detect when the risks are taking place and intervene only for those risks. Instead of over-alerting about each and every problem, Nauto can alert the driver to the highest combination of risks, such as when a driver is tailgating and looking at their phone at the same time. The company uses high-precision algorithms to ensure it’s a high-risk situation before alerting the driver.”
Predictive looks forward about three-to-five seconds into the future, time enough to give the driver coaching or a warning to avoid risk. Because many drivers are unaware that they are even taking a risk, a warning is helpful to allow them to self-correct. Nauto’s goal is to have the drivers self-correct rather than be reprimanded by supervisors.
“In our experience, 80% of drivers self-correct and don’t need supervisor intervention,” says Stefan. “When we turn on feedback, we usually see an 80% drop in risk behaviors in about 72 hours. You do see some drivers that don’t improve at the same rate, so the fleet manager then needs to train and coach those drivers.”
Fleet managers can also use predictive AI to look at the aggregate level for an entire fleet. Using this week’s driving patterns, they can determine what next week or next month’s collisions and risks will look like. They can examine the specific risks to each city they operate in and view patterns of where and when risks happen. By identifying the patterns specific to their fleet, the fleet manager can ensure the safety programs focus on the correct issues.
How Nauto’s AI Solution Meets a Company’s Needs
With a variety of safety technologies to choose from, Stefan recommends that fleet managers begin by identifying the safety problems affecting their fleet and the kind of collisions they’re having. Then, once you’ve identified the problem, you can choose solutions to address the individual problems.
“We tend to see distraction as the biggest problem facing fleets,” he says. “Distraction causes about 60% of the dollar loss across our hundreds of fleets. If this is the case, then a technology with a real-time preventive capability for distraction is essential. You won’t solve your problem by adopting telematics that identify speeding when your problem is really distraction.
“Distraction is a universal problem that is much bigger than most people are aware. Federal data shows that 1,500 fatalities a year are from distraction, but our data that measures distraction in real-time leads them to believe it’s over half of all collisions. That would mean 20,000-25,000 fatalities per year across the US are due to distraction.
“We see distraction as rampant and pervasive for most drivers in most fleets. Speeding, on the other hand, is usually only an issue for 10-15% of drivers in fleets. Only a small set of fleets have big drowsiness issues, usually ones that drive between midnight and 5:00 AM or ones that do heavy manual labor. In a typical fleet there’s only 2-3% of drivers that really suffer from drowsiness in a severe way, and you can then offer very targeted help to those drivers by changing up their route and time of day.”
The Future of Predictive AI
While predictive AI is already capable of providing some valuable insight, there are still a few different growth areas for this technology, says Stefan: “There are still some risks that AI doesn’t yet detect and prevent. This includes things like high speed head-on collisions, which can’t be detected by just cameras alone, but need radar or LIDAR technologies. Another gap is identifying driver medical issues such as a stroke or some other health condition.
“Another unsolved problem is how to make sure that the flow of traffic as a whole is optimized. We’re doing a lot of work to share data around where the risks are with the infrastructure operators, the departments of highway of the cities, and the vehicle makers. Auto and truck makers are becoming more interested in integrating predictive AI into their vehicles, too. We’re also working with a few light- and heavy-duty commercial partners to integrate the capability directly. The benefit for fleets is obvious– you can order a new vehicle with auto pre-installed from day one.”
Stefan concludes that the last area of growth for predictive AI is how it will transform insurance. The company foresees partnerships with insurance companies where fleets will get a discount for adopting the AI capability. In fact, Nauto is close to an opportunity where the insurance will pay to have the predictive AI added to the fleet, allowing fleets to save on both the cost of the equipment and the cost of higher collision rates.
“If you don’t have these systems in the future and other people do, you’ll be at a commercial disadvantage due to higher costs from collisions,” says Stefan. “Your reliability will also be at risk, and you’ll be paying higher insurance rates. It’s a great opportunity not only for commercial fleets, but for anyone living on or near roads to see the number of collisions lowered.”
