AI to stop trains from running into elephants at night may be ready in six months

"We conducted visits to realise the enormity of the problem — along Ghat sections, there are reserve forest areas on both sides of the railway track, and at times, two rail lines run parallel to each other — that doesn’t allow a train driver to be alerted in real-time on elephant movement in the region. So, the only sustainable real-time solution we are left with is AI.”

Some of the basic requirements mentioned by the Tamil Nadu Forest Department in its Request for proposal (RFP) include the need for real-time alerts, predictive technology, night vision and tech-enabled zoning of regions known for elephant movement.

"If there’s an elephant herd already on the track, that’s too late to alter the speed of the train or drive the herd away,” said Supriya.

"Ideally, our AI solution should track herd movement from two kilometres away, which gives us sufficient time to alert the engine driver or forest staff," she said.

To successfully implement the system, the government has proposed dividing the region into colour-coded zones, with red, yellow and green zoning used to denote herd movement in relation to its distance from railway tracks.

"If there’s the danger of herd moving into the red zone, we’d like to know when the animals are in the yellow zone itself," Supriya said.

One of the shortcomings of existing conventional systems is its obvious handicap after dark.

"Most elephant herd movements happen at night," said Supriya. "So, we want a solution that comes with thermal cameras and an AI system that oversees a database of elephant herds and their locations. This lets us create a storyboard of these animals and their movements as opposed to ineffective piecemeal solutions," she added.

The major advantage of using data-enabled AI is that over time, the technology can be utilised to understand elephant behavior better. This could make preventive techniques more effective — given that behavioral patterns are mapped over months, seasons and years.

While 23 IT firms participated in the tender, the Tamil Nadu Forest Department has confirmed that four bids have been finalized. One of these will be selected to build and implement the AI solution.

"One of the conditions that we have put forth in the tender is that the company will have to conduct a field demonstration of the solution, which should satisfy parameters set by a technical committee,” said Supriya Sahu.

This committee is represented by the Wildlife Institute of India, experts from the IT sector and five officials at the state-level including Supriya herself.