Video: Hari Balakrishnan theorizes about predictive analytics and saving lives on the road.
Hari Balakrishnan has some thoughts about road fatalities and how to improve outcomes. They might be especially valuable in the age of the self-driving car.
“This (road fatality problem) is becoming worse,” he said, noting an overall rate of 1.3 fatalities per 100 million miles of road traveled in America, which he said represents a 21% increase since the onset of the pandemic.
Balakrishnan noted three important components to overall risk – driver risk, vehicle risk and road risk.
Stakeholders, he said, tend to know quite a bit about the first two categories. Insurance companies study driver risk, and manufacturers know a lot about the risks of various vehicles. What they don’t know as much about, he suggested, is a more ambiguous type of risk called “road risk” – something that has to do with infrastructure, the context of operating a vehicle around other drivers, and different types of less concrete factors.
Noting that people started commercializing technology on mobile devices about 10 years ago to get better data, he said many researchers are doing a good job on evaluations. But that road risk is still tricky to quantify.
“Imagine you put the world’s safest driver and the world’s best vehicle on our roads,” he said. “There’s an inherent level of extrinsic risk caused by other vehicles, caused by the nature of the road geometry, and topology.”
The goal, he said, is to use high-resolution crash maps to try to aggregate data and predict where crashes are more likely.
Using the phrase ‘kernel density estimation,’ he talked about traffic studies in Los