IIT Guwahati researchers develop traffic models to predict congestion on Indian roads

Researchers at the Indian Institute of Technology (IIT) Guwahati have developed mathematical and machine learning models to predict traffic movement on Indian roads, using drone-based data collected from a highway on the outskirts of Delhi.

The study, led by Professor Akhilesh K Maurya of the Department of Civil Engineering, analysed more than 36,000 vehicles and nearly nine million vehicle positions to understand how traffic density, vehicle speed and spacing change under different road conditions.

The research addresses a limitation in conventional traffic prediction models, many of which are based on relatively orderly road conditions and homogeneous vehicle movement. Such models may not accurately reflect Indian roads, where two-wheelers, three-wheelers, four-wheelers and pedestrians share limited road space, resulting in unpredictable traffic patterns.

To examine these variations, the researchers used drone footage collected during daytime traffic on an eight-lane divided highway near Delhi. The data was used to study the relationship between traffic density and vehicle speed, including variations in speed among vehicles travelling under similar conditions.

The analysis revealed that vehicle spacing changes significantly as traffic density increases. On relatively empty roads, vehicles tend to maintain larger gaps. As traffic becomes moderately heavy, the gaps narrow as vehicles move closer together. However, when traffic approaches severe congestion, the spacing between vehicles increases again, indicating significant changes in driving behaviour.

The researchers also found that nearly 3,000 vehicles could pass a point in each lane under full traffic flow before congestion set in. During traffic jams, when vehicles came to a standstill, density reached nearly 900 vehicles per kilometre across the entire road width.

Using the dataset, the team tested six mathematical models and trained two machine learning models. Unlike conventional approaches that predict a single vehicle speed, the new models estimate a range of possible speeds to account for the variability of traffic on Indian roads.

Explaining the need for India-specific traffic models, Maurya said existing models were developed using traffic conditions that differ from those found in the country. He said drone-based data collection allows researchers to observe large numbers of vehicles and their movements at high spatial and temporal resolution.

The findings could help improve traffic simulation, road and intersection design, congestion management, road capacity assessment and intelligent traffic management systems.

The researchers said collecting similar data across different road types, cities and traffic conditions could further improve traffic prediction in India. Such datasets could also support the development of urban traffic digital twins and adaptive traffic management systems.

The study has been published in the International Journal of Intelligent Transportation Systems Research. It was co-authored by Maurya and research scholars Shekhar Singh and Surya H Ravikumar.