An elephant herd begins its nightly journey across a landscape it has travelled for generations.
Then a train rounds the bend.
Its headlights catch the elephants on the tracks, leaving the loco pilot only seconds to react. Emergency brakes are applied, but a fully loaded train travelling at speed can take a considerable distance to come to a complete halt.
By then, it may already be too late.
This is the problem Indian Railways hopes artificial intelligence can help solve: detect the elephant before the driver can see it.
India is home to an estimated 22,446 wild elephants, nearly 60% of the world’s wild Asian elephant population. The country also has 150 identified elephant corridors connecting habitats across landscapes increasingly crossed by roads, railway lines and settlements.
Railway tracks pose a recurring threat along some of these routes. According to data shared by the Union Government in the Lok Sabha in July 2025, 81 elephants died in train collisions between 2019-20 and 2023-24.
Now, an AI-powered system is attempting to give both elephants and train drivers something crucial: time.
Listening before the collision
Rather than building entirely new infrastructure, Indian Railways developed Gajraj Suraksha, an AI-enabled Intrusion Detection System (IDS) that makes use of the optical fibre cables laid beneath railway tracks for signalling and communication.
The Gajraj Suraksha technology relies on Distributed Acoustic Sensors (DAS), which detect vibrations travelling through fibre optic cables. Every animal creates a distinct movement signature.
Gajraj has been trained to recognise the locomotion pattern of elephants and distinguish it from other ground movements.
Once detected, the system immediately sends alerts to loco pilots, station masters and railway control rooms, allowing trains to slow down or stop before reaching the crossing.
The Gajraj Suraksha technology relies on Distributed Acoustic Sensors (DAS), which detect vibrations travelling through fibre optic cables. Photograph: (Biplab Hazra)
The project, announced by the Railways last year and now being rolled out across nearly 700 kilometres of forest routes, comes at an estimated implementation cost of Rs 181 crore.
According to Indian Railways, the system can detect elephants with 99.5 percent accuracy, identifying movement within five metres of the fibre optic cable.
Instead of asking train drivers to react in impossible situations, the technology attempts to give them time.
Where the risk is highest
India’s rail network intersects several long-established elephant corridors, many of which cut across forests in Assam, West Bengal, Odisha, Kerala, Jharkhand, Chhattisgarh and Tamil Nadu.
Some stretches have become particularly vulnerable.
The Northeast continues to witness a large share of these incidents.
In one of the country’s most widely reported wildlife accidents last year, the Sairang-New Delhi Rajdhani Express struck a herd in Assam’s Hojai district, killing seven elephants instantly.
An injured calf later succumbed to its injuries, taking the death toll to eight.
Such collisions often occur at night, when visibility is poor and herds are moving between feeding grounds.
Why drivers alone cannot prevent them
Following repeated accidents, the All India Loco Running Staff Association (AILRSA), in 2024, publicly highlighted the technical limitations train drivers face.
The association noted that even after emergency brakes are applied, a fully loaded train typically requires at least 1.6 kilometres to come to a complete halt. Whistling and braking, they argued, cannot reliably prevent collisions when elephants appear suddenly on the tracks.
According to data shared by the Union Government in the Lok Sabha in July 2025, 81 elephants died in train collisions between 2019-20 and 2023-24. Photograph: (Biplab Hazra)
Instead, the association called for systemic preventive measures, including AI-powered detection systems, wildlife underpasses and better coordination between railway and forest authorities, rather than expecting drivers to shoulder the responsibility alone.
Gajraj fits squarely within that approach by shifting intervention further upstream, before a collision becomes unavoidable.
Technology works best with people
Artificial intelligence, however, is only one part of the solution.
Tamil Nadu has already demonstrated what happens when technology is combined with coordinated human response.
In the Madukkarai Forest Range near Coimbatore, the Forest Department installed AI-powered thermal camera systems along railway tracks in Puthupathi village. Over the past two-and-a-half years, the network has generated more than 7,116 real-time alerts, leading to 3,281 instances where loco pilots slowed or stopped trains to allow elephants to cross safely.
Officials say the system has enabled approximately 9,481 safe elephant crossings while recording zero elephant deaths on that railway stretch.
The road ahead
Indian Railways has also proposed complementary measures, including solar-powered LED lighting in forest stretches, warning signboards at identified elephant corridors, regular clearance of vegetation near tracks, deployment of elephant trackers by forest departments and honey bee buzzer devices that gently repel elephants away from railway lines.
None of these measures, on their own, can eliminate risk.
But together with AI systems like Gajraj, they point towards a different philosophy of infrastructure, one that recognises wildlife movement as a reality to be anticipated before disaster strikes.
For centuries, elephants have followed the same migratory paths across India’s forests.
They have remarkable spatial memory. Matriarchs lead their herds along migratory routes learned over decades, passing this knowledge from one generation to the next. These pathways connect forests, rivers, feeding grounds and water sources, and many have remained unchanged for centuries.
Railway tracks, roads and other infrastructure arrived much later, and intersected corridors that elephants still instinctively follow. So the jumbos are not choosing danger, they are retracing landscapes they have always known.
If Gajraj succeeds, its contribution may be giving modern infrastructure the ability to recognise these ancient journeys and respond in time.
