As the driver, I am aware that I am supported by some onboard advanced driver-assistance systems (ADAS), giving me greater confidence behind the wheel. This increases my sense of safety and peace of mind.
As I reflect on what’s most important to me in my driving experience – keeping my family safe, I eagerly anticipate the delivery of higher-level ADAS systems and vehicles capable of true autonomy. A system recently announced by Mobileye® that employs Micron memory and storage solutions, demonstrates just how close we are to major paradigm shifts in how we drive.
Preview the future of autonomy with Mobileye
Current cars with ADAS have sensing capabilities that exceed human perception, such as identifying unseen objects on the road and triggering automatic emergency braking (AEB) potentially faster and more persistently than a human driver can. For years, the automotive industry has delivered ADAS for driving automation at level 1 (assistance with individual driving tasks, such as acceleration) and level 2 (assistance with combined driving tasks, such as acceleration and steering) that still rely on the driver to be responsible for control of the vehicle.
However, more sophisticated ADAS features enable a much greater range of uses and thus provide greater opportunities to support the driver in carrying out the responsibility of controlling the car. After many years of ADAS evolving to support more aspects of the vehicle’s operation, we’re now looking ahead from ADAS to autonomous vehicles. We're seeing exciting advancements and even major leaps forward on the road to higher-level autonomous functions.
During the 2021 IAA Mobility show held in September, Mobileye revealed its new self-driving robotaxi where the Mobileye sensor and compute infrastructure is the car’s primary driver. The system is an impactful example of how the future of automotive is shifting. The machine, not the person, will be at the center of driving the vehicle.
What it takes to deliver leading edge autonomous capabilities
An autonomous vehicle’s compute infrastructure could be characterized to accomplish four types of tasks: take data from a variety of sensors (cameras, LIDAR, RADAR, etc.), convert that raw data into a mathematical model of its environment, make strategic decisions about the path the vehicle will take and then execute those decisions by modulating the steering, braking and motor commands.
This is much easier said than done, as higher-level autonomous capabilities require a larger number of sensors and much greater compute performance to perceive what is around the vehicle and then make decisions based on that data. Similarly to other mission-critical intelligent edge applications, autonomous driving can be achieved only by positioning artificial intelligence capabilities nearer to where data is being generated to alleviate the latency caused by sending data to the cloud – in this case, processing data within the vehicle itself. Achieving these capabilities is introducing new demands on automotive compute and memory and storage infrastructures, requiring:
- High data throughput with low power consumption
- Small data package size
- Competitive cost
- Automotive-qualified solutions
- Ability to withstand the rigors of automotive use environments
Micron provides our partners with the compute foundation for autonomous driving
Mobileye supports autonomy requirements through its Mobileye Drive™ self-driving system includes the core advanced hardware AVKIT58. AVKIT58 features 8 EyeQ™ 5H SoCs that collect data from 13 cameras, nine LIDAR1 units and six radar units.
Mobileye selected five (three of which are discussed below) different Micron memory and storage solutions as key elements for its AVKIT58 system:
- 128Mb Quad SPI NOR Flash and 512Mb Xccela NOR Flash - Typically used for reliable storage of boot code, NOR Flash is often the best memory choice for embedded systems due to fast random read performance and low pin count, the latter reducing development cost and the part’s physical footprint. Lauded for its rapid boot performance, Xccela flash is used in applications requiring compute hardware to be ready just a few hundred milliseconds after boot.
- 16Gb and 32Gb LPDDR4 – Micron LPDDR4 automotive qualified solutions provide the memory bandwidth for advanced capabilities like perception operations and data fusion while using less energy per transferred bit than their predecessors2. Micron, within its automotive-grade portfolio, commits to long-term support and the ability for these solutions to operate across extended temperature ranges, which makes these devices excellent choices for automotive applications like Mobileye’s.
- 256GB SSD – Automotive applications like navigation, 3D high-definition maps and data recording (black boxes) demand high-density storage solutions that are resistant to shock and vibrations, support extended temperature ranges and provide data protection capabilities while still providing excellent sequential read/write performance. Micron solutions like those selected by Mobileye fit the bill.
Mobileye plans to deploy its robotaxi for commercial driverless services in Tel Aviv and Munich starting in 2022, operating under the MoovitAV service branding.
Collaboration leads to innovation and greater peace of mind with autonomous vehicles
Micron is grateful to have worked closely with Mobileye for several years, helping achieve the compute capabilities that their solutions require. As Mobileye works toward self-driving capabilities that offer the possibility of safer transportation, less congested roads and improved fuel efficiency, they will bring a tangible and positive impact on our society and economy.
Micron’s work with Mobileye is a reminder that Micron’s DRAM-, NAND- and NOR-based memory and storage solutions enable our customers' autonomous vehicle systems and help enable larger societal benefits to reduce human error and to help keep each other safer on the roads. In the future when we are all driving autonomous vehicles, I look forward to even greater peace of mind for me and even more peaceful dreams for my children.
1 Sensors that work on the same principle as radar but using light from lasers.
2 Data derived from comparing LPDDR4 data sheet to predecessor.