IIT Guwahati uses drones and AI to model Indian traffic flow

Why existing models fall short

Many established models were developed for orderly traffic with fairly uniform vehicles. Indian roads are different: two-wheelers, three-wheelers and four-wheelers share space with pedestrians, which produces greater variability in vehicle interactions, speeds and spacing.

The data showed that on relatively empty roads vehicles keep larger gaps. As traffic becomes moderately heavy, gaps shrink and vehicles follow more closely. In very heavy traffic near congestion, gaps widen again. Maurya said this means the link between density, spacing and speed is not necessarily linear. Before congestion set in, nearly 3,000 vehicles passed a point in each lane under full flow, while in standstill jams density reached nearly 900 vehicles per kilometre across the full road width.

Predicting a range of speeds

The team tested six mathematical models and trained two machine learning models. Rather than output a single speed, the models estimate a range of possible speeds, which the researchers say represents the inherent variability of heterogeneous traffic more realistically.

According to Maurya, such India-specific models could support more realistic traffic simulation, better road and intersection design, congestion management, capacity assessment and intelligent traffic-management systems suited to Indian conditions. Traffic stream models can also inform signal control and emergency response, he said.