Roshy John believes that if you can’t buy, you invent. As the seventh and latest iteration of his artificial-intelligence-driven self-driving car rolled out last October with much promise, the robotics engineer is convinced about AI changing human lives for the better. In Thiruvananthapuram, at a consortium on ‘New approaches to education in the 21st century’ recently, he took a ride to the beginnings of his pioneering work, and spoke about how windows of innovations are being thrown open by AI and robotics.
Roshy says he decided to follow his instincts in 2010. While returning home in a cab from Bengaluru airport late at night, he noticed that the driver was drowsy. Tapping into his expertise, he set out on the task of independently designing what would be India’s ‘first self-driving’ car.
“The idea was irresistible,” says Roshy, now Global Head, Robotics and Cognitive Systems, Tata Consultancy Services, Kochi. “As an engineer, I knew it was an expensive concept. On the other hand, I wanted to find out how it could be actualised and how I could motivate more people to work on this,” he says.
The ‘self-driving’ car fashioned out of a Tata Nano by Roshy and his team
First, Roshy and his 30-member team of robotics engineers prototyped a referential model through 3D simulation.
“We developed the necessary expertise on automobile technology, thanks to advancements in the automotive field. By 2010 or so, motor engines started coming with sensor-based digital systems, moving over mechanical ones,” says Roshy.
One spoke in the wheel was automatic transmission, a technology that was not available then. However, Roshy proceeded with a Tata Nano with a standard petrol engine and manual transmission system. “We designed our own automated manual transmission (AMT),” he says.
Infinite scenarios
The challenges lay in ‘teaching’ the robot, how to drive. “A driver-less car is a large mobile robot that has to work in a highly dynamic and complex environment constituting humans. There are infinite surprises and we had to simulate every possible road scenario and feed them into the AI,” says Roshy, who did his PhD in Robotics from National Institute of Technology (NIT), Trichy. For this, his team utilised machine learning, where thousands of cases or situations are taught to the AI so that it starts learning to read and tackle new ones as they arise. The AI reads its environment through a network of sensors and cameras digitally rigged to it.
Roshy chose a Tata Nano, as the model came with a rear engine, enabling attachment of other moving actuators in the front. Also, its compactness allowed for advantages in computation and cost-cutting. Though the prototype entailed six sensors — two on top, three on the front and one at the rear — and five HDR cameras, Roshy has now managed to bring the costly specs down to two sensors in the seventh and latest iteration.
Currently, the battery-powered innovation has the capability to drive by itself ‘under controlled situations’ at a maximum speed of 50 kmph. It can read its mapped environment up to a radius of 200 metres and as close as to less than a metre. “We are constantly working on improving its computing power. As of now, it stops when there’s a human in front, but it cannot yet read the human, whether it is a child or a man or a pregnant woman. That is where the future work lies,” he says with optimism.
It is not just the ‘chaotic traffic in India’ that proves a poser. Moral, philosophical and legal questions put the brakes on commercial viability, even if one overlooks the massive production cost. Roshy responds to them with a chuckle, by quoting one of Isaac Asimov’s famous laws of robotics, “A robot may not injure a human being or, through inaction, allow a human being to come to harm.”
Nonetheless, he is confident of the power of autonomous cars to help people in the future, especially the differently-abled. “If it works in India, it will work anywhere in the world,” asserts Roshy.
