AI now helping in reducing car accidents in India; Here is how it works

As Artificial Inteligence takes over the world, the emerging technology has been now helping in reduction in road accidents by almost half in a few months. Avneesh Agrawal says, their AI device is having on customer fleets globally. In India, which has the world’s worst road fatalities record, the effect could be dramatic.

Avneesh, who spent years at Qualcomm, including as its research head, and who holds over 200 patents, developed the product together with Qualcomm colleague David Julian as part of a venture called Netradyne that they founded in 2015. Netradyne today has about 800 employees globally, with the majority of its R&D team in India.

The company has over 3,000 customers, all of which are commercial fleets – taxis, limousines, school buses, large inter-state vehicles. Customers are across the US, India, Canada, UK, Australia, New Zealand, and Germany.


How does the device work?
The device integrates multiple sensors – GPS, LTE connectivity, several cameras, inertial sensing unit – and is fitted on the windshield. A software ties it all together.

“The real innovation is in the software where we are running all of the algorithms at the edge and really understanding the world around us,” Avneesh told ToI.

The AI in the device can figure out if the driver is distracted, drowsy, speaking on the phone, following too close to another vehicle, not following traffic signs. And accordingly, it can alert the driver with a beep and even inform the fleet safety manager. As the system tracks a driver’s behaviour over time, it also assigns each driver a green zone score’, an estimate of the driver’s risk of having an accident.

The latest version of the product also has micro weather recognition capabilities. It can detect if there is snow on the road, if there is fog. “So, now we can provide combo alerts, combine the driver behaviour and weather conditions, and accordingly modify the alert definitions,” Avneesh told ToI further.

(with ToI inputs)