Company Says Tiny Motions Can Predict Big Problem: Impaired Driving

While competitors tend to rely on new hardware, CorrActions uses software alone to detect and mitigate impaired driving.

December 20, 2024

Chris Clonts

CorrActions says that it can accurately detect impaired driving with just the real-time data from a vehicle’s steering wheel. It usually uses other data, too, but said the micromotions by impaired-drivers’ hands give away their mental state. Volvo wasn’t mentioned as a manufacturer who will deploy the system, but the company is a CorrActions investor. (Volvo)

A five-year-old Israeli company says it has a software-only, AI-driven solution that can detect impaired drivers more quickly and with less intrusion than interlock devices, passive breath detection or other methods that rely on physical hardware.

CorrActions joins a crowded field of companies, from startups to the biggest global auto suppliers, in the quest to detect and deter impaired driving. The segment looks to be lucrative, since both the National Highway Traffic Safety Administration (NHTSA) and the European New Car Assessment Program (NCAP) have begun exploring regulations requiring vehicles to detect and mitigate the problem.

And a problem it is. According to NHTSA, of 42,000 drunk-driving-related crashes in 2022, more than 13,000 were fatal. And those don’t include incidents caused by drowsy or distracted driving.

CorrActions CEO Ilan Reingold said the company’s R&D team includes scientists with backgrounds in neuroscience, software and AI. (CorrActions)

SAE Media had the opportunity to speak with company CEO Ilan Reingold ahead of CES 2025 in Las Vegas, where CorrActions will demonstrate the technology. Reingold said the company uses data, including steering-wheel and pedal input and in-cabin radar, to detect more than substance-impaired driving. “We're able to accurately detect various cognitive impairments such as alcohol and fatigue and distraction and more, and our road map includes [detecting] stress and medications,” he said.

Reingold was quick to clarify that the system does not compare steering wheel input to ADAS point clouds of the perfect path or anything like that. “Don't think overt motion,“ he said. “We're not looking at how you're driving. We actually filter that out. We're looking at these unconscious, unfelt, submillimetric micro-motions that are caused by the fact that the electric signals that operate our muscles are a reflection of our brain activity.“

Different OEMs would handle warning info differently. Some might try to alert the driver through a display screen or by vibrating the steering wheel. For drowsy driving, if intervention is early enough, it could point out upcoming rest stops where coffee is available. That was the promise of one such impaired driver detection system being developed by Magna. For more serious impairment, CorrActions technology could potentially forc the car to pull over and stop.

The software first filters out unhelpful noise, then uses AI and deep machine learning to determine whether a particular signal set indicates impairment. Reingold said it can account for discrete differences in signals from different drivers.

And just how accurate is CorrActions‘ software? Reingold said that in testing, the system returns less than one false positive for every 350 hours of driving time, which is about how much time an American spends behind the wheel in one year.

“One of the most important KPIs in our solution is the driver's trust in the system,” he said. If the system would pop up false alerts frequently … then drivers will not trust the system and will ignore alerts. And if that's the case, we haven't done anything.”

Reingold said that because the system filters out so much data before any analysis is done, the processing requirements are not heavy. “That means it can run on a central processing unit, but it can also run on linear [edge computing units] and it can also be implemented in vehicles through over-the-air software update and not necessarily during manufacturing,” he said.

The company is already working with OEMs in North America, Europe and elsewhere. He said the first deployment of the system would be in early 2026. CorrActions has raised $10 million since it was founded. Among investors are Goodyear, Volvo and Blackberry.

It’s not just passenger vehicles that could see reduced crashes and fatalities. Fleet management companies have shown extreme interest in the CorrActions system for risk management, though the system there operates slightly differently. “We take the telematics data that's being uploaded from each fleet vehicle today to [the fleet company’s] cloud and we run our software in the cloud and in most of the use cases we report to the fleet manager or to the fleet safety officer, not necessarily to the driver. Maersk Growth, the investment arm of the global shipping giant, is also an investor.

Depending on requirements, the system could even notify insurance carriers and or law-enforcement agencies in something akin to getting a red-light ticket today.

Reingold said the company’s R&D team is skilled primarily in three areas: neuroscience, AI and software.