IIITH team designs AI-solution with 96 pc tracking accuracy

Hyderabad: A professor and a student from International Institute of Information Technology, Hyderabad (IIIT-H), have developed an automated solution for improving air surveillance, which has been deployed at Bharat Electronics Limited (BEL).

While the existing tracking mechanism used by the Indian Air Force was providing an accuracy of 91 per cent, the AI-assisted tracker created by the IIIT-H team has demonstrated 96 per cent accuracy, according to a press release.

In defence forces, all flying objects in Indian airspace, including potential threats and commercial aircraft, are detected and tracked by ground-based radars produced by BEL. The current system used by the IAF relies on a multi-sensor tracking (MST) mechanism via radars located at different parts of the country.

With overlapping radars sensing sometimes the same aircraft, and delay in communication between sensors, two common errors emerge. One is the 'merging' error — where multiple aircraft in close proximity to each other are incorrectly identified as one and 'splitting' — where a single aircraft is sensed as multiple and erroneously flagged as a threat, the press release said.

To address these issues while retaining the original tracking mechanism, a team from BEL Ghaziabad approached IIIT-H for development of an automated solution.

The IIIT-H research team, led by Prof Paruchuri with Masters student Anoop Dasika, trained a machine learning model with 11 days' worth of anonymised and labelled data collected from 17 million data points captured by various radars. The algorithm that they developed resolves the issues of merging and splitting errors, leading to a five per cent increase in accuracy. Another advantage is that it helps in a detailed radar analysis.

"By identifying which radars are contributing to the maximum number of errors, and which error in particular, such a probe helps in prioritising replacement or repair of the concerned radars," said Prof. Paruchuri.

The software has been transitioned to BEL and is currently being tested out in their simulation environment.