IIT Guwahati tracked 36,000 vehicles by drone to predict traffic jams before they form

By TUF News Desk · Published 6 November 2026 · 4 min read

Fact-checked against 2+ independent sources · Editorial policy

Nearly 3,000 vehicles passed a single point in each lane every hour before the road seized up. That figure, measured from the air above a highway in North Delhi, now sits inside a prediction system built at the Indian Institute of Technology Guwahati.

Researchers at the institute have developed AI-based models that predict how traffic behaves on Indian roads. The system combines high-resolution drone footage with mathematical modelling and machine learning, and it is designed for roads where cars, two-wheelers, three-wheelers and larger vehicles share the same space, often alongside pedestrians.

Why Indian roads break most traffic models

Traffic forecasting tools built for other countries usually assume vehicles move in orderly lanes at similar speeds. Indian roads rarely work that way. Differences in vehicle size, acceleration and driving behaviour create interactions that those models miss, so their predictions fall apart in heavy, mixed traffic.

The Guwahati team set out to measure Indian traffic as it actually moves. They flew drones over an eight-lane divided highway with wide paved shoulders in North Delhi and tracked individual vehicles continuously. The dataset covered more than 36,000 vehicles and nearly 9 million vehicle positions.

Research scholar Shekhar Singh said drone-based collection lets the team watch large numbers of vehicles over long stretches of road and follow each one without interruption. That kind of detail is hard to get from roadside sensors, which usually record only vehicle counts and average speeds.

Six maths models, two machine learning models

The team tested six mathematical models and trained two machine learning models on the drone data. Instead of predicting one speed for a given traffic density, the models estimate a range of possible speeds, which reflects how much speeds vary between vehicles travelling in the same conditions.

The study also produced two concrete measurements of how Indian traffic behaves near capacity. Before congestion set in, nearly 3,000 vehicles passed a point in each lane under full traffic flow. Once vehicles were virtually stationary in a jam, density reached nearly 900 vehicles per kilometre of road across its entire width.

Professor Akhilesh Kumar Maurya of the Department of Civil Engineering, who led the work, said the models could shift traffic management from reacting to congestion after it happens to predicting bottlenecks before they develop. The research has been published in the International Journal of Intelligent Transportation Systems Research.

What traffic authorities could do with it

The models give planners a way to test ideas before spending money on them. Traffic engineers could simulate how different vehicle categories interact at different densities, check whether a road or junction has enough capacity, and judge whether widening a road or redesigning a junction would actually improve flow.

In cities, the same approach could support adaptive traffic signals that change green-light durations based on live conditions rather than fixed cycles. It could also help authorities spot stretches where traffic is turning unstable, forecast congestion before a jam forms, and work out the likely effect of a diversion or a change in signal timings.

Maurya said India-specific models could support more realistic traffic simulation, better road and intersection design, congestion management and capacity assessment. As cities build digital twins of their road networks, he added, high-resolution traffic data paired with AI prediction could become a central part of them.

Students considering IIT Guwahati for civil engineering or AI research can read more about the institute on the official IIT Guwahati website. Our guide on how to choose a university in India sets out the trade-offs worth checking before picking an institute.

Frequently asked questions

What did IIT Guwahati researchers build?

AI-based models that predict traffic behaviour on Indian roads, using high-resolution drone footage combined with mathematical modelling and machine learning.

How much traffic data did the study use?

More than 36,000 vehicles and nearly 9 million vehicle positions, recorded by drones above an eight-lane divided highway in North Delhi.

Where was the research published?

In the International Journal of Intelligent Transportation Systems Research.

What can the models be used for?

Traffic simulation, road and intersection design, signal timing, emergency management and intelligent transport systems that respond to live road conditions.

Frequently asked questions

Sources: Business Standard, Press Trust of India