India’s roads pose a huge challenge for self-driving car technology. Despite unique traffic rules on its roads, from darting pedestrians to stray cattle, Indian auto giants are determined to get autonomous vehicles out on the streets.


As the rest of the world rushes into self-driving car technology, India, soon to be the world’s third largest auto market, has barely got started.

Amid perpetual traffic gridlock and its legion of lawbreakers, from pedestrians darting through traffic to stray cattle on the roads, Indian automakers, startups, and engineering schools are now slowly steering the development of self-driving cars for their roads.

India’s roads present a true deep learning challenge. The trick is to teach the machine the unspoken rules that govern traffic there.

Researchers suggest that India will have to do this very differently from anywhere else on the planet.

At MCity, a test track for driverless cars at the University of Michigan, cars go through simulated environments with the aim of training these vehicles to circumvent obstacles and follow road signs. “Copying and pasting this technology from the West” would not work in India, according to Alec Gallimore, dean of engineering at the University of Michigan.

For one, India’s traffic is unpredictable, with a lack of signage and lanes merging at random. Then there’s this whole other problem – driverless cars are banned on Indian roads. To work around this, startups and auto giants – like Tata Motors and Mahindra – implement a dual system in these autonomous cars where drivers can override the system and gain full control.

Here’s an alternative plan for India: stick to trials in gated communities like hospitals and college campuses. Autonomous vehicles here will travel at much slower speeds and in far less traffic, before being tested in real-world conditions along fixed lanes at designated times.

But even then, large-scale use of autonomous vehicles has its own set of challenges. India’s weak internet and frequent power outages pose a huge problem for smooth rides.

There are two ways driverless cars could work around those issues.

One, they could download data and maps and then analyze the differences from what they see through their cameras and sensors, making use of inputs from sensors on the road to guide and direct the vehicle.

Two, autonomous vehicles could communicate with each other at a distance up to 450 meters (1,500 feet). The cars can gather data from each other on velocity, direction, and more. In doing so, the vehicle builds up a detailed picture of obstacles to avoid.

Here’s where it gets tricky: finding the best model to work with terrible internet.

Alec Gallimore, dean of engineering at the University of Michigan, suggests a combination of the two: tweaking the combination based on the strength and latency of the network in each particular region in India.

There’s also a unique cultural and economic barrier to consider. In India, there are around 5 million domestic staff, some of whom serve as private drivers. The low cost of such staff might discourage India’s middle class from upgrading to an autonomous car.