In the autonomous vehicles (AVs) industry, engineers, data scientists, and business leaders have all identified true self-driving vehicles as not just a priority, but an eventuality. That narrative is so strong, however, it’s easy to forget that the mass deployment of AVs still faces significant challenges that extend beyond training AI.
Every step of autonomous vehicle development has been met with obstacles surrounding data, whether it be training machine learning models or reading/writing data quick enough for vehicles to respond. Engineers must work to balance the rewards that assistive AV technologies can offer today with the data challenges of the AV future.
Data traffic control
One of the chief issues facing AVs today is how and when to send and receive data. Engineers are working to understand the role of data based on when and where it needs to be shared between vehicles, the cloud, and edge data centers. In vehicles, many different sensors and technologies create data, but much of the heavy work within AVs is done via computer vision.
“Vehicles have cameras and devices to sense the world around them, mostly for safety purposes,” said Russell Ruben, automotive segment marketing director at Western Digital. “Some of that data is processed in real time and then discarded. Other data is captured to learn and train AI, so it’s then offloaded, sent to the manufacturer, and then deleted locally. It’s a cycle of continued learning and improvement.” When data isn’t quickly accessible, though, vehicles need to reach out and find new data to inform their decisions.
V2X, or vehicle-to-everything, communication is one of the first methods for vehicles to gain those new insights. V2X leverages the advancements in telecommunications to send and receive critical information between a car and road infrastructure, other cars, pedestrians, or just about anything that falls into the Internet of Things.
V2X is considered an important piece in the emergent data communication puzzle for AV developers because it helps vehicles understand what’s around them and how to respond. The quick data exchanges between the vehicle and other entities are akin to a human driver looking in a mirror, assessing the situation, and adjusting their speed or changing lanes — mundane but vital tasks for expert drivers.
“Vehicles need to communicate with one another on the road, which is where V2X comes in,” said Matteo Zammattio, a flash storage field applications engineer at Western Digital who works with carmakers. “The cars use 5G and telecommunication technologies to tell other vehicles its own properties, and roads infrastructure as well, like streetlights. Companies believe V2X communications [are] the major key-turn solution.”
Infrastructure is a great example of how V2X might work in the near future. The two entities would use the technology to share data, passing intelligence about journey length or average wait times at intersections. This data could then be collected, sent to a data center, and then used by traffic management systems to improve traffic patterns.
While not as fast as V2X, the edge is an additional option for data communication, which offers a larger data set and computation power to the vehicle, but with lower latency than a central cloud computing center. While the cloud might offer more resources, the closest data center might be miles away, not nearly fast enough for a car to react safely.
“With 5G latency, we’re talking 10 milliseconds, edge computing in 10s of microseconds, and communicating with the cloud around 25-30 milliseconds,” said Zammattio. “But decisions are taking place instantaneously.”
Storage, on and offboard
Getting all the data on the vehicle’s storage requires impressive write performance and accessing that data demands read performance to match. Crafting storage solutions that are equal to the task is as vital as implementing latency-busting 5G and edge data centers.
While specifications and form factors are subject to change, the one constant for automotive data storage is flash. Due to the volatile environment of a car, resilient silicon wafers are the preferred solution. Their reliability enables auto manufacturers to rely on their performance when it matters most.
“Flash is the choice for data storage in cars today,” Ruben said. “[Auto manufacturers] take a NAND device, for example, and they want to solder the drives to the board so it’s more resistant to shocks.”
Beyond the environment, the storage on a vehicle is expected to go through a lot of churn. While cars today collect a bevy of data, much of it is discarded once it’s no longer useful. Hyundai estimated that their AV prototypes generate 10GB of data every second, but only a subsection of that is stored. Flash, again, is the preferred solution for constant writing and rewriting.
High performance — reliable, consistent, and rapid storage — in such a difficult environment is why Western Digital is in constant communication with car manufacturers to understand and develop solutions that meet their high standards.
“A big focus for automakers is safety,” said Ruben, “the additional safety that comes from the data and the technology that enables them.”
Collaboration and standardization
Between data communication, data storage, form factor, and regulation, there are several groups that have a stake in how AVs are deployed. A simple right-hand turn in a city might leverage telecommunications, cloud architects, hardware and firmware engineers, and machine learning specialists.
Precisely because of the complex collaboration between all these systems is why communication and standardization are vital to the success of AVs. Western Digital is part of the ecosystem of organizations that sets standards and ensures requirements are met in new storage products. Western Digital is in conversation with automakers to communicate the power of new technologies and how they will impact data storage and transmittal in vehicles. Simultaneously, auto manufacturers are communicating their needs for safety and performance, so the storage industry can rigorously design and develop solutions.
While setting industry standards doesn’t make headlines, it’s what brings together the legacy of collaboration in this space. From the moment an autonomous vehicle leaves the lot, it will constantly orchestrate the dissemination of data from system to system. Data makes it possible, but it’s communication and collaboration that will make AV’s engineering marvels a reality.