Former Uber CEO Travis Kalanick famously said, after experiencing New Delhi’s chaotic roads, that India will be the last place in the world to get self-driving cars. But a handful of startups think the country could be the perfect testbed for creating autonomous vehicles that can handle anything.
A recent video from Bhopal-based startup Swaayatt Robots suggests they’re making progress. In the 6-minute long clip, a sensor-laden SUV weaves through narrow unmarked streets, dodging pedestrians, dogs, cows, slow-moving tractors, and a constant stream of scooters overtaking, cutting across, and even driving on the wrong side of the road.
Swaayatt CEO Sanjeev Sharma says the video highlights the two major characteristics of Indian traffic that makes it so challenging. It is both stochastic and adversarial, which in simpler terms means that road conditions and driver behavior are almost entirely unpredictable, and that other road users are more likely to play chicken than give way.
While self-driving cars developed by Western technology companies have already begun commercial operations, this rollout has been made possible only by training on millions of miles of driving data painstakingly gathered over many years. And despite all that training, these companies are still bedeviled by the “long tail problem”—the idea that no matter how many scenarios you train on, you will always encounter rare but unique “corner cases” that will flummox your vehicle. Here’s where technology developed in India has a major advantage, says Sharma.
“The kind of traffic and environment we’re negotiating, the entire navigation course can be labeled as a corner case,” he says. “This is the most complex that it can get for an autonomous vehicle. If you’re able to build here, this technology is universal.”
Tackling India’s uniquely unruly streets requires a different approach to that taken in the West, says Sharma. Cars made by companies like Waymo and Cruise are loaded with sensors, including cameras, radar, lidar, and high-precision GPS, and they rely heavily on high-definition 3D maps. Their goal is to create a highly detailed and deterministic model of the environment around the car, Sharma says.
But while that might work on the orderly, gridlike streets of Phoenix, it won’t get you far in India. As a result, the Swaayatt team has gone back to the fundamentals to create algorithms that generate probabilistic representations of the environment. This process is normally very computationally expensive, but Sharma says they’ve found a way to do it in real time, though he’s cagey about the details. Broadly, the approach relies on “data-efficient reinforcement learning” but also involves modules that uses game theory to model interactions between different road users as well as computer-vision systems that predict where absent or faded lane markings should be to help the car navigate.
While Swaayatt’s SUV does feature a suite of sensors, including lidar and high-precision GPS, these sensors are mainly used to gather training data, says Sharma. In the company’s demo video, the vehicle is using nothing more than off-the-shelf cameras.
India’s roads force a new approach
India’s challenging road conditions force engineers to take more inventive approaches to self-driving, says Gagandeep Reehal, cofounder and CEO of Bengaluru-based self-driving startup Minus Zero. Founded in 2021, the company has the ambitious goal of building an “Android for self-driving cars.”
Reehal is also adamant that the approach taken by larger, Western companies won’t translate to India. As well as being reliant on large numbers of sensors and high-definition mapping, these companies take a modular approach to self-driving, breaking up the problem into multiple smaller tasks like object detection, localization, and motion planning, each served by their own algorithms.
Minus Zero is instead designing a single holistic system capable of creating a “world model” that builds in a more general understanding of physics and road-user behavior. Reehal says the overarching goal is similar to that of Canadian self-driving startup Waabi and United Kingdom–based Wayve, which are also focused on building end-to-end models rather than modular ones, though their approaches differ in the details.
Rather than using standard deep learning to blindly search for patterns in large amounts of driving data, Reehal says Minus Zero has created physics-aware algorithms that can more readily pick out the most salient information, even when trained on smaller datasets. The company also relies on multiagent learning approaches that better capture the complex interactions between different road users.
This more general approach to self-driving is crucial when designing for India, says Reehal. “When people tried to solve autonomy in structured countries like the U.S. or the U.K., it was their natural tendency to make the model learn the rules,” he says. “The moment you come to a country like India, you realize you’re really have to solve generalized locomotion.”
Minus Zero’s near-term goals are relatively modest, focusing on developing an AI “copilot” that can take over some highway driving. The company recently signed a multiyear partnership with Indian truck maker Ashok Leyland to develop the technology, which Reehal predicts could be on the road within the next three to four years. But he says Minus Zero is also working with international automotive firms, and he agrees with Sharma that if Indian companies can solve self-driving at home, they can solve it for the world.
How an Indian self-driving pioneer sees the market
Few people have been pushing to bring self-driving vehicles to India’s streets for as long as Roshy John. In 2016, while working as the head of robotics and cognitive systems for Indian IT giant Tata Consultancy Services, he modified a Tata Nano hatchback to drive autonomously. He continued to work on the project, but the Indian market was not ready for the technology at the time, he says.
That situation is starting to change, he says, thanks to the country’s rapid development and greater government support for technology startups. In 2021, John quit his job at TCS to found the self-driving startup RoshAi. Thanks to his early start in the field, John says his company is sitting on a mountain of Indian driving data, which it’s using to build a highly customizable self-driving solution to sell to automotive companies and equipment makers.
The company’s software is similar to the modular systems developed by companies like Waymo, but John says this structure makes it easier to tailor the software to the needs of individual customers. While RoshAi provides fully autonomous systems that make use of a full suite of sensors, its technology can also be adapted to provide simpler driver-assistance capabilities that typically rely on just cameras and radar. “India is a cost-conscious market,” John says. “Most of the automotive companies want to go for much cheaper solutions.”
RoshAi’s offerings go beyond software. If you want to cede control of a vehicle to a computer, a key component is a drive-by-wire (DBW) system, which replaces a car’s mechanical controls with electronic ones. John’s company sells a DBW system that can be retrofitted to almost any vehicle, and he counts a host of automotive firms and self-driving startups among his customers. “I’m helping other companies to fast track their development in driverless systems too,” he says.
Will full autonomy take off in India?
While Indian self-driving cars are still firmly in the testing phase, high-end Indian-made cars are already beginning to feature simpler driver-assistance systems, including technology made by German engineering giant Bosch.
Sandeep Nelamangala, president of Bosch Mobility India, says these systems typically fall into two categories. The first effectively acts as a second pair of eyes for the driver, providing things like automatic emergency braking and lane departure warnings. The second provides partially automated driving on highways that allows drivers to take their hands off the wheel in certain situations.
These products are built for a global audience, but Nelamangala says they’ve had to adapt them significantly to take account of Indian road conditions. He’s skeptical of the argument that autonomy technology developed in India will easily translate to other countries.
“It cannot be simply assumed that the success of these systems here shall guarantee applicability worldwide,” he says. “Every region has its own set of unique challenges like public infrastructure, driving laws, motor-vehicle usage patterns, topography, climatic conditions, and so on that need to be analyzed and addressed independently.”
It’s also not clear if there’s really a market for true self-driving technology in India, says Vinay Piparsania, founder of MillenStrat Advisory & Research. When you can hire a full-time driver for less than US $150 a month, the capital outlay on a self-driving vehicle could be hard to justify, he says.
Safety is likely to provide a more compelling argument, he says, considering the country recorded more than 160,000 road deaths in 2022. But for the technology to take off, there will need to be a concerted build out of high-quality road infrastructure, Piparsania says, and a greater focus on standardization to ensure that things like signage and road markings are more consistent.
That’s why he thinks that although India may well end up a leader in the development of self-driving technology, it’s unlikely to benefit much from it in the near term. “We have the talent to do these things, but we may not have the environment where such applications could be used at this moment,” he says.