Waymo 研究：完全清醒司机夜间致命车祸风险仍达白天3.1至3.9倍

Imagine a world where we successfully remove every single drunk driver from the road. Achieving this would be a monumental victory. Alcohol impairment remains a massive, critical crisis on public roads, directly contributing to over 30% of all traffic deaths in the U.S. Eliminating it is a mission we proudly support alongside long-standing partners like Mothers Against Drunk Driving (MADD).

But our latest peer-reviewed research highlights a second, equally sobering truth that we cannot ignore: traffic fatalities rarely result from a single factor, and improving road safety requires examining conditions beyond driver behavior alone.

This is where the broader potential of autonomous technology emerges. Beyond introducing a driver that is always sober, Waymo offers a unique opportunity to address the wider, interconnected risks of our roadway ecosystem.

The Sober Driver Baseline

To understand how deeply embedded roadway danger really is, our first study — accepted for publication in the Journal of Safety Research — set out to answer a question: How well do humans perform when they are completely sober?

Because national traffic databases do not track vehicle miles traveled (VMT) by a driver's blood alcohol concentration (BAC), our researchers built an innovative "exposure reconstruction" model. By cross-referencing toxicological data from the Fatality Analysis Reporting System (FARS) with real-world telematics, we estimated the miles driven by sober (zero BAC) versus impaired human drivers across the 50 most populous U.S. urban areas. We then used these figures to establish unimpaired driving baselines in each city. This spatial and temporal mapping is a major breakthrough for public safety research, offering unprecedented visibility into local impaired driving patterns.

Building on our previous studies that established local human benchmarks across different road types and times of day, this new zero BAC cohort represents a deeper layer of research granularity. It offers an opportunity to evaluate the safety of autonomous vehicles against a higher safety bar — a non-impaired human driver.

The results show that removing alcohol-involved drivers from the data pool reduces the average urban fatal crash involvement rate by 23%—dropping from 1.42 to 1.10 incidents per 100 million miles. The impact of alcohol varies heavily by geography. In Houston, for example, removing impaired drivers lowers the fatal crash rate by a staggering 31%. In contrast, in Minneapolis, the reduction is 13%. This underlines why it is important to use local human benchmarks rather than national averages when evaluating road safety.

MADD CEO Stacey D. Stewart said: “Waymo’s work here can fill a major gap in road safety research. Finding a way to granularly track where and when alcohol-impaired driving happens gives the safety community a much clearer picture of where the real risks lie on our roads. This underscores why cutting-edge autonomous technology can be a vital tool — alongside anti-drunk driving technology built into all human-driven vehicles — to protect lives where human limits meet the road.”

Crucially, the study evaluated sober driving risk across different times of day. One hypothesis may be that elevated nighttime crash rates are solely due to impaired driving. However, our study indicates that even for 100% sober drivers, nighttime fatal crash risk remains 3.1 to 3.9 times higher than daytime risk.

This is partly because sober drivers still need to react to impaired drivers. However, this persistent, elevated risk overnight suggests there are other systemic hazards. Reduced visibility, inadequate lighting, driver fatigue, structural speeding layouts, and the necessity of interacting with other unpredictable road users are known nighttime risk factors that create a highly volatile domain that alcohol sobriety alone cannot correct.

The Intersection of Compounding Risks

To explore these hazards further, our second study utilizes the U.S. Department of Transportation’s Safe System Approach. This framework recognizes that while humans inevitably make mistakes while driving, those mistakes should not cost them their lives. A safe roadway relies on redundant layers of protection: Safer People, Safer Vehicles, Safer Speeds, and Safer Roads.

Our paper, published by the International Research Council on the Biomechanics of Injury (IRCOBI), evaluated 37,654 real-world fatal cases from the 2023 FARS database, revealing a stark reality about how risks overlap on public roads.

Unsurprisingly, elements that featured heavily in this fatal dataset were impairment, speeding, and seat belt/helmet misuse or nonuse. Over 77% of all fatal cases involved at least one of these three elements and 9.2% of fatal cases involving all three. But there are a multitude of factors that influence a crash: crucially, only 8.2% of fatal cases involved strictly human behavior factors alone. The overwhelming majority represented a fatal convergence: for example, an unlit road layout together with a massive vehicle size mismatch, combined with unexpected high speeds.

Changing the Status Quo

An idealized human who never drinks, speeds, gets distracted, or gets tired is a driver that simply doesn't exist. Waymo’s goal is to reduce traffic injuries and fatalities in the cities we serve. So to measure our true safety impact, we benchmark Waymo’s performance against real-world human driving under actual road conditions. Establishing a sober driver baseline gives us an additional research lens to evaluate our capabilities against an optimized human standard.

The Safe System approach is built on the core principle of shared responsibility — recognizing that road safety is not the burden of any single actor, but a collective system of safe road users, safe vehicles, safe speeds, and safe roads working together. By designing the Waymo Driver to respect all elements of the Safe System — constant 360 vigilance, following posted speed limits, defensive driving, and zero impairment or distraction — we can break the chain of overlapping risk factors. In fact, our analysis estimates that autonomous systems could address up to 83% of the system failures contributing to passenger-vehicle-involved fatal crashes across the United States.

Mauricio Pena, Waymo’s Chief Safety Officer said: “To truly understand road safety—and to properly evaluate the impact of autonomous vehicles—we must recognize that traffic fatalities are a complex, systemic issue. Fatal crashes typically result from compounding factors—like low visibility, high speeds, and poor road design—that even sober drivers cannot always overcome. The Waymo Driver is built to solve for both sides of that equation, eliminating the impairment factors entirely while simultaneously accounting for environmental vulnerabilities and addressing systemic hazards that contribute to traffic fatalities.”