Flash Memory Making Bid in Data Storage Wars

Connected and autonomous vehicles of the near future will be running a suite of complex processes and applications that require robust and scalable flash memory solutions.

With the swiftly accelerating volume of data comes the need for faster processors and more memory capacity. Also the move towards advanced driver assistance systems (ADAS) will further heighten requirements for even more memory intensive, multi-layer high definition maps. Voice recognition and speech synthesis applications will be party to this trend too particularly because there are a number of cars that already support natural language speech recognition.

“Flash memory is the only economical way of storing large amounts of data in the vehicle on a solid state device,” Ian Riches, executive director of global automotive practice for Strategy Analytics, told TU-Automotive. He explained that, for autonomous vehicles, the need to store sensor data from radar, camera and LiDAR in some form of “black box” data recorder, although not yet mandated, is seen by pretty much all players as vital to ensure that an automaker can explain the ‘choices’ its vehicle made in the event of any accident. “This data, if deemed to be interesting by the on-board algorithms, can also be sent off-board for further processing and analysis,” he explained. “Such memory solutions could be many hundreds of gigabytes or even terabytes in size.”

Riches also noted that, for connected cars, additional storage to help facilitate over the air (OTA) software updates is likely to be another area of growth with the major automotive NAND flash players including Micron, SanDisk/Western Digital and Toshiba, among other major global players. “Obviously, as flash costs fall it will allow OEMs to offer more on-board storage,” Riches said. “We are not seeing a competitive technology to flash at the moment for on-board mass storage of data.”

Complex functions in current cars already have to manage a various amount of data, from sensors, navigation and other components. Rising connectivity and access to clouds enable new customer functions and services, such as function on demand or OTA updates and, therefore, additional flash storage capacities for applications and regarding backup partitions is needed.

Bernhard Augustin, development domain controller for Audi’s autonomous driving division, explained that because in the past the main requirement in control units was the ability to execute simple instructions, there was almost no need for flash storage at all. “As the amount of data to be handled constantly increases, microprocessors with complete operating systems and file systems are necessary,” he said. “Furthermore legal requirements, for example concerning a black box recorder in automatically driving cars, will increase the number of storage in vehicles.”

He also noted ADAS require short boot times for safety critical systems, as well as a high write performance and lifetime considerations. “ADAS will use real time information from back-ends or clouds, for example concerning hazards, traffic light phases,” Augustin explained. “To provide that information during operation in real time, data have to be cashed in working memory (RAM).” In addition, some data additionally has to be quickly available after a system restart, like the exact localization of a parking car (particularly related to parking infrastructure), which he said is a good example of a vehicle requiring sufficient flash memory capabilities.

Niclas Nygren, senior director of vehicle software and electronics for Volvo, noted memory products have become more advanced in terms of the their capability to hand errors or wear and tear of the memory but there is no silver bullet solution yet. “You have a limited amount of usage in terms of write and that’s why the more advanced memories have automatic correction and blocking of memory, where you can get help from that memory component to make sure you write on the good memory,” he said. “Any operation of the car requires you to have all components running correctly and we need to design the cars so that any failure, should it occur, has a path of handling any errors from any component.”

He also explained that advanced mapping and navigation features would require more robust memory solutions, as would sensors and applications constantly monitoring the health of the vehicle. “In general, you need flash memory in all applications in the car as such, whether it be simple software updates, the storage of not only software and program code but also databases like maps, which need to be stored locally,” Nygren explained. “There’s not one set of specifications that covers all scenarios.”

He noted today’s manufacturers have a lot of computers in the car and some of them have very read/write intense functions, such as for the powertrain, whereas map data is updated less frequently. “We need to find the right specifications for each domain,” he said. “That means pooling the resources, so you have a lot of memory in one centralized computer, which means additional questions of reliability and durability from that perspective.”

— Nathan Eddy is a filmmaker and freelance journalist based in Berlin. Follow him on Twitter.


Leave a comment

Your email address will not be published. Required fields are marked *