GPUs Revolutionising The Driverless Automotive Ecosystem

GPUs Revolutionising The Driverless Automotive Ecosystem


SUNDARA R NAGALINGAM is Head – Manufacturing and Energy Industries at NVIDIA Graphics Pvt Ltd

Technology today touches every aspect of our lives, and the reasons are simple – convenience and experience. Whether it is about controlling your AC through your smartphone or simply making a Skype video call, these applications improve our everyday lives. The technologies that have made our phones smart and our computers super are converging on the automobile.

Driving this technological upsurge now are cars. Autonomous cars represent a dominant and unstoppable trend in the motor industry. For users, this will make driving safer, more convenient and less stressful than ever before.

For automakers, this transformation poses significant challenges. Data gathering and analysis, artificial intelligence and the speed of innovation are some of the new areas on the OEM's agenda. Safety must continue to be primary to the carmaker's approach, but it will also require a robust technology strategy – one that will ultimately lead to self-driving cars. The powerful and versatile nature of graphics processing units (GPU) is making autonomous vehicles a reality.


What happens when you give wheels to a supercomputer? A robot that can take you anywhere you wish. Similarly, imagine you could call your car from its parking to pick you up. How the world would change if your car could follow all the traffic regulations without touching the steering wheel. This will become possible because these vehicles are gaining the competency to sense and understand their surroundings.

Majority of accidents are instigated and caused by humans. Over a million people worldwide are killed in auto accidents every year, the vast majority caused by driver error. Ill-judged manoeuvers, inattention and speeding are some of the most common factors responsible for road accidents. Such problems dissolve, when the driver is a machine that can never be side-tracked and receives more information about its surroundings than a human could ever process.

The technology behind driverless cars will address other such issues as well. Fewer hold-ups caused by accidents and more predictable driving behaviour will streamline congestion and fuel efficiency. And there will be new solutions to the perennial parking problem. When your car can automatically drop you off at your endpoint, then park itself, it wouldn't matter where the parking area is located.

For many consumers, the idea of extensive ADAS (advanced driver assistance systems) features or a fully self-driving car is fraught with safety concerns. However, the truth is that improving safety is the number one reason why autonomous vehicles are such an important trend. Volvo has long been synonymous with safety. Now, the Swedish automaker's latest efforts could make self-driving cars synonymous with safety, too. Henrik Lilnk, Senior Technical Leader, Volvo Car Group says, "Nobody should be killed or seriously injured in a new Volvo car. Long term, new Volvos shouldn't crash."


But a new generation of super-smart cars requires serious Artificial Intelligence (AI). There aren't enough engineers in Silicon Valley to hand-code software to account for everything that happens when you drive. To deal with all the stuff a car sees on the road, you need a new kind of technology. This technology is called 'Deep learning' (DL), which is at the core of the autonomous car's development.

As a concept, DL has been around for several decades but was too computationally demanding to be practical. A part of AI, DL allows computers to perform tasks for which they have not been specifically programmed. Now, thanks again to a processor called the GPU (graphics processing units), originally developed to power video games, deep learning is accelerated for real-world deployment. It is revolutionary because it lets computers teach themselves about the world through a training process that's roughly similar to the way children learn. Large amounts of training data are fed through sophisticated algorithms to build Deep Neural Networks (DNN), which are complex mathematical models that mimic the human brain.

The DL approach is so powerful that last year GPU-powered deep learning systems exceeded human levels of perception for the first time. GPUs are ideal for digesting the massive amounts of information involved in deep neural networks. They have already accelerated the training of deep neural networks by 20 to 30 times over the past few years. This trained model turns out to be the brain of the autonomous car. It is constantly alert, has faultless reaction times and it never rests from learning. All that sums up to the formula needed for the auto industry's ambitious goals for zero road accidents in the future.

Certainly, there are good reasons why automakers traditionally run on very long design cycle timelines. Your car is a life-critical device and needs to go through exceedingly rigorous testing and certification to ensure it is up to the task. However, by maintaining these vital standards, while calling on the agile expertise of technology companies, car manufacturers are proving that it is possible to achieve a happy medium, where innovation and safety go together.

Powered by deep learning on GPUs, a system of automated electric vehicles, known as WEpods, just made history in the Netherlands by becoming the first vehicles in the world without a steering wheel to be given license plates. Driverless cars are already a reality in Japan and will transport sports persons to venues during the 2020 Tokyo Olympics. In Europe, Belgium, France and Italy have already started planning their transport systems around this. London is not far behind, with autonomous pods set to form the epicentre for public transport in the city. They are already in use at Heathrow airport.


We will see this hitting the Indian roads in the next few years. However, considering the enormous amount of vehicles on the roads and poor infrastructure, it will take time for all the cars to become autonomous. Let's say the roadmap for autonomous cars in India will be similar to revolution of mobile phones in 1990s. The change will be gradual, but definitely ground breaking.