America's roadgoing drunks still account for a third of all traffic deaths despite years of public outcry and shame. Now researchers say vehicles have enough technology to sense an impaired driver at the wheel — with the right software.
This isn't the much touted alcohol-sensing technology under development by several companies and pushed by federal regulators to measure whether a driver has any alcohol in their blood. Instead, the team of researchers from several universities and the National Highway Traffic Safety Administration asked a different question: Could software using the sensors such as stability control and lane departure warnings already in some vehicles deduce whether a driver was drunk or impaired based solely on how they drove?
Using a University of Iowa driving simulator, the team had 108 volunteers with blood alcohol levels of up to 0.10% drive two courses, then monitor their responses. Sifting through the data and applying three algorithms, the team found that even their simple first stabs could predict about 80% of drivers with alcohol levels above the legal maximum of 0.08%, in as few as eight minutes of driving — regardless of the age or gender of the driver. The standard field sobriety test given by police officers has about the same accuracy.
The research comes with caveats, the largest being the most obvious that a drunk driver can do a lot of damage in eight minutes behind the wheel. The study looked at moderate drinkers; alcoholics may not be so obvious. And the most accurate sensor of drunk driving — lane departure — is also the one most prone to problems operating in bad weather or visibility.
But the research could have uses beyond rum-soaked rallyers; drivers can be impaired by drugs, fatigue or other reasons that blowing on a tube won't catch. You can read the executive summary here.