Putting trust in the driverless seat: How cloud and AI can help consumers believe in AVs
A hybrid cloud approach is key to combining the billions of gigabytes of data required for autonomous vehicles
It’s a rite of passage.
The moment when a parent tosses their keys to their newly permitted teenage driver and takes the long and trepidatious trip into the passenger seat. Like a song drifting between stations you no longer control, feelings of sheer and utter terror mix with the future promise of some well-earned free time.
Freedom, as they say, doesn’t come easy.
Now imagine that same mix of disaster-may-strike fear and how-did-I-ever-live-without-this excitement every single time you got into the car.
That’s exactly the feeling auto manufacturers are hoping consumers can overcome when they finally get behind the non-wheel of autonomous vehicles. To get there, automakers need to deploy some of the most advanced technology ever developed. They also must prove to consumers they can trust that technology—and not just trust it to get them from point A to point B but trust it with their very lives and property.
To build that trust, car companies are relying on petabytes of information, millions of miles and hours of testing data, generated by development fleets of autonomous vehicles roaming the world to ensure their systems work flawlessly.
To Frank Kramer, a client technical architect at IBM Systems and expert in advanced driver assistance systems, or ADAS for short, data management is one of the greatest technical challenges ever attempted on our streets.
“AI is based on data and statistics, and to verify that the statistic is correct we have to deal with triple-digit petabytes” of real-time information, Kramer said during a panel at this year’s IAA auto show in Munich. “This is pretty new to all of these players in this game, and verification and validation is the important thing to make sure that these systems will be safe on the road, and they are insurable.”
Putting the pedal to the metal
The insurance industry is a central part of the trust component as carriers will be responsible for underwriting the policies of millions of autonomous vehicles that will one day be on the road. They are also seen by regulatory bodies and consumers as an objective third party on the safety of these systems because of their financial culpability. Without their support, AVs are unlikely to ever hit the road.
This makes finding the clues to how autonomous systems respond in real-world driving situations critical—whether it be a stop sign, a road hazard or a person crossing the road. “Probably the most important scene is just a few seconds” Kramer said.
Locating that sliver of insight requires that every data point is organized and available, including car, road, sensor, weather, buildings and landscape among thousands of known factors. Data scientists and engineers need access to an incomprehensible amount to data—at least incomprehensible to humans—in order to train and build the technology to understand how specific parts of the autonomous system are performing.
This is where AI and cloud become essential tools to their work, and particularly a hybrid cloud approach that can accept inputs from almost any system. Hybrid platforms like Red Hat OpenShift allow teams around the world to collaborate on this massive challenge and accelerate the development process. “It’s not a single point problem, with big data it’s a distributed, big problem around the world and a very, very compute intensive one,” said Frank.
Putting the pedal to the metal is critical for automobile companies to compete with fast-moving IT companies new to the automotive category but whose business is data.
Virtual simulations are one way to do that. “You cannot change the weather on the road as you probably need it for your testing, in a virtual world you can very easily change the reality,” Frank said. “This multiplies your data in a very, very efficient way,” added Mr. Frank.
Seamless orchestration on an open hybrid cloud platform
The challenge may appear daunting, but the automotive industry has proven time and again to be one of the most adaptive industries on the planet. From mass production to quality control to greater safety standards and emissions regulations, the automotive industry knows how to innovate. The list of safety advancements alone is impressive: three-point seatbelts, airbags, ABS brakes, crumple zones to today’s advanced driver assistance systems paving the way for fully autonomous vehicles.
It’s a race that will serve multiple players–new entrants and established auto manufacturers alike–but most importantly drivers, their passengers, and pedestrians, who will benefit from the advanced safety systems autonomous vehicles are built upon. Think about the possibilities when you take drunk, distracted, and drowsy motorists out of the picture. Autonomous driving has tremendous potential—but with great potential comes great responsibility.
“We have to come together as a team,” Frank argues “Nobody is able to solve this problem alone, so you have to bring together the specialists–the automotive specialists, the AI specialists, the hardware, the networking specialists, the data specialists, and the cloud technology to make it a package that is flexible, scalable, cost-efficient that can work with this huge amount of data to get this problem solved.”
And you have to bring everyone together in a way that works, adjusts, and adapts. Data and real-world events are constantly changing. And just as on the roads, one mistake or modification in the data can jam things up. Seamless, reliable orchestration on an open, hybrid cloud platform is the surest way to smooth driving and the necessary trust that comes with that.