Today, if we are the lawabiding kind, we often have to wait at signals even if there’s no traffic flow from the other side crossing us. That’s because signals have manually-fixed timers.
Globally, AI-based systems are emerging, which together with cameras, can figure out what the optimal signal time should be, depending on traffic conditions, and change the signals accordingly. But building such systems is extremely complex, given also the differences in things like traffic patterns and driver behaviour across cities and countries.
Ashish Dhamaniya, professor at Sardar Vallabhbhai National Institute of Technology, Surat, together with his PhD student Rajesh Chouhan, has been working to build a dataset that captures traffic patterns in Indian cities to solve traffic issues through automated systems. They have used cameras and drones across several cities including Ahmedabad, Delhi, Chandigarh, Surat, Rajkot, Jaipur, and parts of Uttarakhand, and analysed the traffic patterns and the movements of different kinds of vehicles and pedestrians using an AI-based system that they call Shiv-Natraj. The dataset today con tains over 6.5 TB of data on different vehicles at traffic junctions, highways, roundabouts, and toll-plazas.

They are now using this dataset for a variety of use cases related to traffic management. “We have got assignments from different agencies like Surat Municipal Corporation, Ahmedabad Municipal Corporation, NHAI,” Dhamaniya said at the Dadra & Nagar Haveli and Daman & Diu edition of Technovate for India, a joint initiative by TOI and startup ecosystem builder Talrop. If you today have to wait for 200 seconds at a traffic stop, the AI-based system could bring it down to just 60 seconds, depending on its analysis of the movement at that junction, as also movements in adjacent junctions.

The team, Dhamaniya said, is now working to capture driver behaviour through an eye-tracker provided to drivers. This will provide data on whether drivers are distracted, and by what. The eyetracker will also provide other road data that drones cannot capture. Dhamaniya said they will integrate these datasets to enable safer driving. India, he noted, has one of the worst road traffic accident records in the world. He said the system they are developing would be able to predict what would be the optimum speed for different segments of roads, and help bring in regulations or advisories for drivers to reduce issues like overspeeding.