After witnessing the chaotic roadways of New Delhi, former Uber CEO Travis Kalanick is renowned for stating that India will be the last country in the world to have self-driving cars. However, a few firms believe the nation would provide an ideal testing ground for self-driving cars that can tackle any situation.
The Bhopal-based business Swaayatt Robots appears to be moving forward based on a new video. The 6-minute video shows a sensor-equipped SUV swerving around tiny, unmarked streets while avoiding dogs, cows, people, slow-moving tractors, and an endless stream of scooters that are passing, crossing, and sometimes even driving on the wrong side of the road.
According to Sanjeev Sharma, CEO of Swaayatt, the film focuses on the two main factors that make traffic in India so difficult. Because it is both hostile and stochastic, to put it simply, other road users are more inclined to play chicken than to yield, and road conditions and driver behavior are virtually completely unpredictable.
Although self-driving cars created by Western technology corporations have already started to operate in the market, their deployment has only been made feasible by training on millions of miles of meticulously accumulated driving data over many years. And even with all that training, these businesses are still plagued by the “long tail problem,” which states that no matter how many scenarios you practice, your vehicle will eventually run into uncommon but peculiar “corner cases” that will confuse it. According to Sharma, this is one area in which Indian technology has a significant advantage.
The entire navigation path can be referred to as a “corner case” given the type of traffic and environment they are handling, he claims. “For an autonomous vehicle, this is the most complicated situation possible. Universal if you can construct here, is this technology.
According to Sharma, dealing with India’s distinctively disorderly streets necessitates a distinct strategy from that used in the West. High-precision GPS, radar, lidar, cameras, and other sensors are all standard on cars manufactured by companies like Waymo and Cruise, and they also largely rely on high-definition 3D maps. According to Sharma, their objective is to build a deterministic and incredibly precise model of the surroundings of the car.
That could be effective in Phoenix’s well-organized, gridlike streets, but it won’t cut it in India. To construct algorithms that provide probabilistic representations of the world, the Swaayatt team has therefore gone back to the basics. Sharma claims they have figured out a way to perform this ordinarily very computationally expensive procedure in real time, though he is mum about the specifics. In general, the method is based on “data-efficient reinforcement learning,” but it also incorporates computer vision systems that forecast the location of missing or faded lane markings to aid in vehicle navigation, as well as modules that apply game theory to simulate interactions between various road users.
Although Swaayatt’s SUV has a number of sensors, such as lidar and high-precision GPS, Sharma claims that these are primarily used to collect training data. The car only uses store-bought cameras in the company’s demonstration film.
India’s roadways demand a different strategy
According to Gagandeep Reehal, cofounder and CEO of Bengaluru-based self-driving firm Minus Zero, India’s difficult road conditions need engineers to be more creative in their approaches to self-driving technology. The corporation was established in 2021 with the audacious objective of creating a “Android for self-driving cars.”
Reehal is also sure that bigger, Western corporations’ strategies won’t work in India. These businesses approach self-driving in a modular fashion, dividing the problem into several smaller jobs like object detection, localization, and motion planning, each handled by a separate algorithm. They also rely heavily on a large number of sensors and high-definition mapping.
Rather, Minus Zero is developing a single, comprehensive system that can provide a “world model” that incorporates a broader comprehension of physics and the behavior of drivers. Though their methods vary in specifics, Reehal claims that the overall objective is comparable to that of Waabi, a self-driving business located in Canada, and Wayve, a company based in the United Kingdom. Both companies are concentrated on creating end-to-end models rather than modular ones.
Reehal claims that Minus Zero has developed physics-aware algorithms that are more capable of identifying the most important information, even when trained on smaller datasets, as opposed to employing traditional deep learning to haphazardly look for patterns in vast volumes of driving data. In order to more accurately represent the intricate relationships between various road users, the organization also uses multiagent learning techniques.
According to Reehal, it is imperative to have a broader approach to self-driving design while creating for India. According to him, “people’s natural tendency was to make the model learn the rules when they tried to solve autonomy in structured countries like the U.S. or the U.K.” “Upon arriving in a nation such as India, one quickly realizes that generalized locomotion is a real challenge.”
The short-term objectives of Minus Zero are somewhat low-key; they center on creating an AI “copilot” that is capable of handling some highway driving. In order to develop the technology, the company has entered into a multiyear agreement with Ashok Leyland, an Indian truck manufacturer. Reehal estimates that the technology might be deployed in the next three to four years. However, he notes that Minus Zero is also collaborating with foreign automakers, and he concurs with Sharma that Indian businesses can tackle self-driving technology globally if they can handle it at home.
An Indian pioneer in self-driving cars views the market
Few have campaigned as long as Roshy John to introduce autonomous cars to Indian roads. He converted a Tata Nano hatchback to drive itself in 2016 while serving as the head of robotics and cognitive systems at the massive Indian IT company Tata Consultancy Services. Although he persisted in working on the idea, he claims that the Indian market was not yet prepared for the technology.
He claims that because of the nation’s quick development and increased government backing for tech firms, that situation is beginning to change. John left his position at TCS in 2021 to found RoshAi, a self-driving business. John claims that because of his early entry into the industry, his business has access to a wealth of Indian driving data, which it is utilizing to develop a highly configurable self-driving system that it will market to manufacturers of machinery and automobiles.
Although the company’s software is comparable to the modular systems created by businesses like Waymo, John claims that because of its structure, it is simpler to modify the program to meet the demands of specific clients. RoshAi offers completely autonomous systems that utilize a variety of sensors, but its technology may also be modified to offer more straightforward driver-assistance features that usually only require radar and cameras. John claims that the Indian market is price-conscious. “The majority of automakers prefer to use far less expensive solutions.”
RoshAi offers more than just software. Drive-by-wire (DBW) systems are essential for giving a computer control over a car since they swap out the mechanical controls with electrical ones. John’s business offers a DBW system that can be installed in practically any car, and among his clients are numerous automakers and self-driving companies. He claims, “I’m also assisting other businesses in expediting their development of driverless systems.”
Will India adopt complete autonomy?
High-end Indian-made vehicles are starting to be equipped with more basic driver-assistance systems, such as those manufactured by German engineering behemoth Bosch, even though autonomous vehicles in India are still very much in the testing stage.
According to Sandeep Nelamangala, president of Bosch Mobility India, there are two general types of these systems. The first one serves as an efficient substitute for the driver’s second set of eyes by offering features like lane departure alerts and automated emergency braking. In certain circumstances, the second option allows drivers to take their hands off the wheel thanks to partially automated driving on highways.
Nelamangala claims that although these items are designed for a worldwide market, they have undergone substantial adaptation to cater for the circumstances of Indian roads. He doesn’t believe the claim that autonomy technology created in India can be simply applied to other nations.
“It cannot be taken for granted that the triumph of these systems in this location will ensure their universal relevance,” he states. “Every region has distinct challenges that must be analyzed and addressed independently, such as public infrastructure, traffic laws, motor vehicle usage patterns, topography, and climate conditions.”
The market for actual self-driving technology in India is similarly unclear, according to Vinay Piparsania, founder of MillenStrat Advisory & Research. According to him, it might be difficult to justify the initial investment in a self-driving car when you can hire a full-time driver for less than $150 USD per month.
According to him, safety will probably make a stronger case given that the nation saw more than 160,000 traffic fatalities in 2022. However, Piparsania notes that in order for the technology to become widely used, there will need to be a deliberate effort to develop high-quality road infrastructure, as well as a stronger emphasis on standardization to guarantee that elements like road markings and signage are more uniform.
For this reason, he believes that while India might become a leader in the field of developing autonomous technology, it is unlikely to reap significant benefits from it in the near future. “We may not currently have the environment where such applications could be used, but we have the talent to do these things,” he argues.
