You may not need to take the wheel of your car in 20 years, but Big Data will. As we head towards an autonomous future for motoring, human input will gradually be replaced by IT systems. At the heart of these systems will be massive amounts of data, constantly analysing, estimating and assessing the current situation on the road and forecasting future risks.
Self-driving car technology needs to tackle the three Ps: perception, prediction and planning. The biggest hurdle remains prediction. It’s not necessarily what other autonomous vehicles will do, but rather what other human road users – or animals – will do. One way to tackle this is deep learning – or high-performance pattern matching.
All this processing means a mountain of data, transferring from each car. If you consider the microchip shortages at present causing a supply crisis across the auto industry, think of what’s to come.
It doesn’t take an IT degree to realise that data management, processing and storage is going to be a major global issue as we move towards self-driving vehicles.
According to Jeff Fochtman, senior vice president at data storage giant Seagate: "Storage is definitely a bottleneck. As IoT comes online and more information is created by sensors, the gravity of that data is overwhelming.
“I really think of technology as three real components: compute, networking and storage, and they all need to evolve together.”
Test phase
Seagate has developed a new in-car storage device, Lyve, about the size of a shoebox installed in autonomous test vehicles. It handles the streams of data from these vehicles, which can be as large as 150 terabytes (TB) each day. To put that into some perspective – that's the equivalent of 32,000 DVDs. Or to put it another way, the entire US Library of Congress was estimated to contain about 74TB of data back in 2009.
According to Fochtman, during the test phase for autonomous cars, the focus is about collecting as much data as possible and not deleting it or not making decisions. “It’s about gathering data, ingesting that data, creating clouds to build the neural network that the AI runs on.”
He estimates we will be in this test phase for the next 10 to 15 years. “Even as autonomous levels are reached, the test phase which is just continuing to gather all the data. Sometimes there’s 30-40 sensors, including seven cameras, three LIDAR units. It’s just spinning off so much data and no one wants to throw anything away as it is building up. The data is the competitive advantage as we’re getting into regulatory modes.”
Real-world data
That's why Seagate is a key partner in the recently launched Future Mobility Campus Ireland (FMCI). Based in the Shannon Free Zone in Co Clare, it offers companies like Seagate, Cisco, Red Hat and Jaguar Land Rover the chance to test autonomous vehicles – both on the ground and in the air, in real-world settings.
The test facility consists of a road network that is retrofitted with interconnected state-of-the art sensing and telecommunication technologies. Jaguar has already been testing its autonomous I-Pace out of the facility.
Previously, such tests were carried out in facilities in sparse and sunny climates like Arizona in the United States, but pitting these vehicles against Irish weather and road conditions delivers far more real-world data, according to speakers at a recent FMCI congress.
All this data feeds the development of programmes to tackle scenarios faced by future autonomous vehicles as we move towards handing over control of our cars to the computers. Before then we need to manage how it is gathered, and work out a way to create more chips to keep the regular cars on the road, never mind future autonomous ones.
