Meet this IIT engineer developing structured self-driving technology for unstructured Indian roads

Sanjeev Sharma with his self driven Mahindra Bolero

There is no doubt that the interest around self driving technology has escalated like never before. With many big players like Google and Tesla entering the market, names like Tata Elxsi too have been trying hard to get India’s own driverless car. But given the complex traffic conditions and the unstructured Indian roads, the journey is still a far stretched dream.

However, this IIT Roorkee alumnus isn’t losing hope easily. Having founded Swaayatt Robots, he is aiming to create a completely different technology for driverless cars to operate seamlessly in India. Sanjeev Sharma, Founder, Swaayatt robot shares “the navigation module by Swaayatt is designed such so as to handle previously unseen environments. Meaning that it can work in GPS denied environments and can produce high speed trajectories even in very cluttered environments”.

Let’s explore the journey of Swaayatt Robots, currently a one-man army, that promises to bring affordable self-driving vehicles in India.

India vs others-

It is a known fact that the environment in developed countries such as the US is very well structured with well defined traffic rules, which helps to significantly reduce the complexity of traffic-dynamics as compared to India. “The structured roads and environmental conditions such as paved roads, proper & elaborate lanes, road markers and signs makes it easier for vehicle’s perception modules”, says Sharma, who has been working on this idea since 2009.

An evident absence of these scenarios in India requires a careful designing, new perception modules, robust motion planner, and a robust lane markers detector that can detect lane markers on road. Explaining the system developed by Swayaatt, he says “it finds all the divisions of the road, and not just the lane markers pertaining to the current lane of the vehicle”. He also believes that if not for this module, the only alternative way of doing this would be to build a detailed 3D map of every road in the country and then hope that the vehicle’s localization module is robust enough to localize the vehicle on road with 5-10 cm of accuracy.

Sharma shares “currently I am using a Mahindra Bolero to test the technology. It was successfully tested in a colony/campus where I live on February 17th this year. In this video, the entire vehicle is navigating autonomously using just 2 cameras”. However, it still needs some working to make it robust enough for typical Indian city traffic conditions, where there are virtually no traffic rules.

Though the government regulation has been relaxed for self driving vehicle testing, he sees Indian traffic and unstructured condition, a major obstacle. “A very unique motion planning algorithm needs to be in place—such as those which can take decisions such as overtaking a vehicle on typical 2 lane roads in India or crossing a busy intersection at peak traffic hours without traffic signal, which is the case often in India”, he shares.

Further challenges in adoption-

Given that the concept of self driving is still nascent in India, how successful would the efforts by Swaayatt Robots prove? Sharma is quick to answer that he is quite hopeful and optimistic about the future of self-driving vehicles in India. “I had been dreaming of a self-driving vehicle since 2009 and thus, at least to me, the idea of self-driving vehicles anywhere is not as nascent”, he says.

His hope and passion for the idea of self driving cars is evident from the fact that he decided to decline PhD offer at University of Massachusetts, and come back to India to start with his own venture. Though the exact launch date for the technology is under wraps, Sharma reveals that some time in June, it could work on highways. “Working with a complex dynamical system such as a self-driving vehicle is not an easy task, given no support and no funding. But I can assure you that India will have self-driving vehicles in 2017”, he says.

Adding a note on adoption, he says that it will depend largely on safety and prices at which it would be offered. Making a technology that is both highly robust and cost-effective would mean a lot of tweaking. He is already working in that direction by making perception possible by using cameras instead of LIDARS, which typically cost $50000. “My most recent work provided an alternative to LIDAR and it will cost just $1000”, says Sharma who got fascinated to self-driving vehicles during his undergrad, when he came across a video by MIT team in DARPA Grand and Urban Challenges.

Since then he has been gathering all the advanced mathematical knowledge from undergrad along with a know-how on AI, machine learning, data mining, numerical optimization through books and research papers.

And finally the project from his fifth semester on autonomous navigation and motion planning kick started the idea of founding Swaayatt Robots. “This project which continued in the form of theoretical research until August 2013, now plays an important role in the self-driving technology I am developing now”, he shares.

Funding-

Sharma has been bootstrapping Swaayatt Robots for more than 30 months now and once funded, he believes that the technology can be retrofitted on aftermarket products.

“Whether funded or not, I will finish the proof-of-concept and demo it working on highways, campuses and in Bhopal city, in at least on three different routes. However, if I get funding, I will be able to do it a lot faster and a lot more effectively”.

Swaayatt Robots recently got accepted into NVIDIA’s Inception Program, and they are considering the application for the NVIDIA GPU Ventures Program.

Roadmap for the company-

Sharma has set his focus very clear where he aims to have a proof-of-concept (POC) for campuses, highways and cities ready within 5 weeks, 2 months, and 4.5 months respectively.

He has also been working on developing an advanced driver assistance system (ADAS), which is specifically tuned for Indian conditions and would be ready by the end of June and demoed in July. “By December I also want this technology to be tuned for selected few Indian cities like Bhopal and Bangalore, and few others. The plan is to be technologically capable of starting self-driving taxi services in these few selected cities, on selected routes (if unfunded), by December”, he shares.

That’s not all, he also has plans to use this technology for Indian Defense, where a robot with intelligence and autonomous navigation capabilities can help soldiers. He plans to become a global player in the field of AI and robotics.

Concluding note-

He concludes by leaving a suggestion to the government of India that they should revise custom duties on equipment and hardware, that could help pure R&D startups like Swaayatt Robots in the long run. He also adds “Indian government should introduce a law that penalizes self-driving vehicle companies if their vehicle / technology makes a mistake and it results in a fatal accident—if the fault is of the self-driving software”.

Comments

comments

Srishti Deoras
Srishti Deoras

Srishti currently works as Senior Content Strategist for Analytics India Magazine. When not covering the analytics news, editing and writing articles, she could be found reading or capturing thoughts into pictures.