DRIVERLESS CARS FACE UNIQUE CHALLENGES ON INDIA’S CHAOTIC ROADS.
AUTOS
• At a secret testing track outside Bangalore, an arm of the Tata conglomerate is recreating the jumble of Indian roads to develop an autonomous driving system.
That means accounting for pedestrians darting through traffic, multiple lanes that merge without warning, poor signage and stray cattle on the roadside.
India’s push into the driverless race is being driven by conglomerates such as the Tata Group and Mahindra Group along with a slew of startups and engineering schools, which are taking on global giants in an industry Intel projects will spur US$7 trillion of spending by 2050. The country, forecast to soon be the world’s third- largest auto market, is loath to be left behind even as its chaotic roads and regulations create unique hurdles.
“Indian roads present a true deep learning challenge,” said Roshy John, a 17- year veteran in robotics, who heads that business at Asia’s largest IT services provider Tata Consultancy Services Ltd. John’s innovations are riding on sister company Tata Motors’ US$3,500 Nano, touted as one of the world’s cheapest cars, while another unit of the group, Tata Elxsi, is developing the driverless platform.
Since 2013, John’s obsession has been to road test the Nano, retro- fitted with driverless technologies including sensors, actuators and cameras as well as a robotic system to handle the steering, gas pedal and brakes.
But with road rules forbidding autonomous cars on the country’s streets, the vehicle still has all the original systems in place so an alert driver can grab the controls. Without the dual systems, even a YouTube video of a test drive is enough to get John in legal trouble.
Apart from the street anarchy, developers in India have also given up a head start to global giants pouring billions into developing platforms. While Alphabet Inc., Uber Technologies Inc., Ford Motor Co., Baidu Inc. and Tesla Inc. have all invested heavily, nobody has yet developed a roadworthy autonomous solution.
Even if t he r ules are changed to allow driverless vehicles, integrating them into city traffic will remain especially complicated in India. Roads can vary between modern highways and dirt tracks, with erratic street signage, a wide variety of vehicles and the occasional elephant or camel. The ubiquitous, three- wheeled auto- rickshaw is rigged in so many different ways by its drivers that sensors struggle to identify its form, creating complications for machinelearning algorithms.
“After training and feeding hundreds of photos, our system cannot identify 15 per cent of the vehicles on the Indian road,” said Nitin Pai, senior vice- president and head of strategy and marketing at Tata Elxsi. “The driverless car is ready for the road. But is the road ready for the car?”