The advent of artificial intelligence (AI) tools in truck cabs marks the latest wave in the “digitalization” of freight vehicles, joining a lineup that includes video-only dashboard cameras and electronic logging devices (ELDs). But while those previous innovations have had fairly straightforward missions—video-only dashcams record vehicle accidents, while ELDs track driving hours—AI technology comes in many different flavors and can be used to achieve a wider variety of goals.
Those could include analyzing road conditions ahead, assessing driver behaviors, and providing collision alerts. But regardless of how they plan to apply the technology, fleet managers considering AI for their trucks need to understand what it is and how it works in order to select the right system.
That’s not always easy. “There are 250 ELD companies out there, but they basically all do the same thing—maybe some just make the user interface easier to use—because the capabilities are government mandated. But AI is the Wild West, because there’s no mandate. So it’s apples and oranges, [and] it’s really hard for a fleet to dig through it all [to figure out], What is this technology really doing?” says Stefan Heck, founder and CEO of Nauto, a California-based provider of advanced driver assistance system (ADAS) and driver management system (DMS) technology.
To make that determination, it helps to know a little bit about how the technology works. Installed on a tractor-trailer, an AI dashcam is a smartphone-sized box attached to the windshield about where you’d put your toll transponder. The box contains chips for processing and data storage, a forward-looking digital camera, and often a driver-facing camera as well. Many are also linked to cameras in the truck’s cargo area or rear end, or to a telematics device that records how fast the operator drives, how hard they brake, and so forth.
Typical AI dashcams measure all those variables multiple times per second and synthesize the results into a single, digital worldview. The unit then wrestles the data through proprietary algorithms to assess road risks in real time: Is there a car in the road ahead? How far away is it? Is this a close-following situation? Is that in the parameters of what we consider tailgating? If so, should we notify the driver and ask him to increase his following distance? Or is the driver’s foot already on the brake pedal, so an alert would be redundant?
Ideally, a real-time AI dashcam acts like a cool-headed coach who quietly corrects only the most serious errors, as opposed to a backseat driver who nitpicks the driver’s every move.
IS YOUR HEAD IN THE CLOUDS?
Given all the market confusion, how do you find the right “coach” for your operation? As always, the answer depends on what you’re looking for. But if, like many, you’re looking for the kind of real-time alerts described above, one of the key things to find out is where the AI processing is taking place—that is, is it occurring on board the truck or on a cloud computing platform in another location?
That’s an important distinction, Heck explains. If the algorithms run on an in-cab device, the AI can analyze road risks nearly instantaneously and provide collision-avoidance coaching in real time. But if the system relies on remote processing, time lags come into play, which means it can only analyze events after the fact—what Heck calls “better than nothing”—but can’t support truly real-time analysis of driving patterns as they happen, he explains.
Another important consideration in selecting an AI dashcam is accuracy. That might seem like an obvious point for anyone who’s purchased consumer electronics or office equipment lately, but the stakes are higher with vehicle technologies. In the case of AI dashcams, accuracy problems could cause the unit to send too many or too few alarms. While too many alarms might sound more like an inconvenience than a major problem, that’s not the case, according to Heck. “The fewer alerts the better,” he says, “because people get [ticked] off with too many alerts. If you get four out of five false alarms, you’ll start tuning them out. And some in-cab warning systems have a 40 to 50% accuracy rate, so drivers will ignore it because it’s wrong half the time.”
Like Heck, Barrett Young considers accuracy in flagging risky driving behavior to be a key differentiator in the AI dashcam market. “If a driver is alerted for something they’re not actually doing wrong, then the driver doesn’t trust the camera, and they’ll [end up having] awkward conversations with their fleet manager,” says Young, who is chief marketing officer at Netradyne, a California-based developer of fleet safety solutions that says Amazon is its largest customer. “And if your manager is constantly slapping your hand for doing little things wrong, then that relationship is not going to be very good,” he explains.
One way around that problem is to use the dashcam not just to track drivers’ transgressions but also to reward positive driving behavior. Netradyne uses inside-the-cab alerts it calls “micro-coaching” to change behaviors like seatbelt noncompliance, following vehicles too closely, or texting while driving. But it also awards “driver stars” to those who use a defensive driving maneuver to reduce risk, for example. Some fleets have developed rewards programs based on those stars, handing out bonuses or giving extra time off to their top-performing drivers.
IS THE AI DASHCAM YOUR FRIEND?
As for the economics of outfitting a fleet with cameras, AI dashcams typically generate a quick return on investment (ROI) through savings on fuel consumption, maintenance costs, and insurance premiums, says Abishek Gupta, VP for product management at Motive, the California-based fleet technology company formerly known as KeepTruckin. (Among other channels, the firm provides its AI-powered dashcam solution in partnership with Platform Science, a company that provides mobile devices for commercial fleets.) Those savings could come by discouraging drivers from behaviors like rolling stops, distracted driving, sudden accelerations, or tailgating, for example.
But to achieve the best results, fleets need to prove to their drivers that AI dashcams are accurate, trustworthy, and working to support them, not spy on them. “The accuracy piece has to work because your driver has to trust it. If he can’t trust it, he won’t listen to it,” Gupta says.
Then there are the privacy concerns. While some warn that truck drivers will quit their jobs rather than submit to high-tech surveillance, Motive has found that this claim is not supported by statistics, Gupta says. “People think if they install AI dashcams, their drivers will all leave. But whether they have no camera, a road-facing camera, or a driver-facing camera, we have seen almost no change in driver retention rates. Still, it’s important to [incorporate] education and enablement in training to get buy-in before you just roll it out.”
In the end, the best way to pick the right AI dashcam for your fleet is to try them out yourself, Gupta says. To get a real feel for what each system can do, he says, you have to obtain test units from various vendors, install them on different fleet vehicles, and compare the results over time.
