Cookie Information

We use cookies to improve the website experience of our users. To learn more about our cookies, visit our Cookies Policy page.

logo-micron

Add Bookmark(s)


To:

Email


Bookmark(s) shared successfully!

Please provide at least one email address.

Microchip background

Autonomous Driving Systems Drive Increased Memory Needs

Autonomous Driving Systems Drive Increased Memory Needs

Automobile manufacturers already use memory in a variety of applications: infotainment; cluster/dashboard; Communications and advanced driver assist systems (ADAS).  Currently these systems use DDR2, DDR3, and LPDDR2 to meet their memory needs. As automotive OEMs and their customers drive the advancement of connectivity in cars and continue to develop additional safety features, the triangulation of need, technology and timing is driving the automotive adoption of new technologies like LPDDR4, e.MMC 5.x, Octal NOR; and new form factors like multi-chip packages (MCPs) to solve their needs.  In this blog, we discuss the memory changes occurring to support advancement in ADAS Systems.

Recently the National Transportation Safety Board (NTSB) called for all cars in the US to have collision avoidance systems, expanding ADAS systems and implementing Vehicle to Vehicle (V2X) systems.  Micron predicts a move from single system cameras, like back up cameras, to more advanced multi-purpose, multi-camera systems enabling safety features like surround view camera, lane departure warning and more.  This would allow ADAS to move beyond warning the driver that they are leaving the lane, to the vehicle actually helping steering correct or in the more advanced systems the vehicle will apply the brakes to potentially prevent an accident all together.  In the future we also see the ADAS system moving from the individual camera modules performing the processing to a centralized system that will effectively be the brains of the car.

The modern ADAS system that we describe above consists of multiple sensors, for example, high resolution/high frame rate cameras, radars, laser scanners, and night vision. These sensors, especially the video cameras, drive high data bandwidth that is written into a volatile memory. A DSP (digital signal processor) unit reads the sensors data from the memory and runs a complex ADAS algorithm. That algorithm performs multiple memory accesses for each video frame and demands high memory bandwidth.  This process often becomes the ADAS system performance bottleneck with current systems, which is why an LPDDR4 solution is recommended. It supplies the high bandwidth the ADAS system demands.

Architecture Bus Width @ Bandwidth / pin Total Bandwidth
LPDDR4 x32@3200MT 12.8GB/s
DDR4 x16@2400MT 4.8GB/s

Besides bandwidth, another major concern for automotive manufacturers is size.  They need memory configurations with the right mix of density, power, performance, temperature, reliability, cost, and support. Multi Chip Packages (MCPs) are being used to support active safety systems while greatly reducing the memory footprint. Memory MCPs stack non-volatile memory (NVM) (which delivers boot-up/application, operating system, and other critical code/data execution) and volatile memory (RAM) (which serves as high-speed temporary memory) together in one package, reducing area footprint.  In addition to reducing area footprint, MCPs also have a lower ball count and increased performance and density. MCPs ease design considerations by offloading the embedded memory of a microcontroller unit (MCU) using industry standard JEDEC interfaces and memory types.

For critically fast response time, many existing sensor modules that employ high-performance and high-density LPDDR are moving to more centralized NOR + LPDDR2 MCP solutions to align with complexity, intelligence, and density requirements.

The next step to provide even more safety on the road is for ADAS designers is to create an integrated system network that will interact with other cars and with roadside traffic monitoring stations. By exchanging anonymous, vehicle-based data regarding position, speed, and location, vehicle to- vehicle (V2V) communication enables a vehicle to detect other vehicles with a 360-degree awareness of their positions and any threats or hazards they present. It then calculates the risk and issues the driver warnings or takes preemptive actions to avoid or mitigate crashes.

The next generation self-driving cars on the road will need  Vehicle-to-Vehicle (V2V) and Vehicle-to-infrastructure (V2I) communication more commonly known as V2X communication.  V2X is the wireless exchange of critical safety and operational data between vehicles and roadway infrastructure. With V2X communication enabled, a car becomes autonomously aware of surrounding traffic conditions – even when its driver is distracted. Like with V2V technology, V2I can suggest a corrective action to the user or even take full control of the situation limiting the risk of an accident.  Along with microprocessors and computing power, memory is a key element to the success of these applications. Connected to V2X requirement is the essential need for data security; recent successful attempts to hack cars through their electronic systems have drawn everybody’s attention to this topic.  Look for future blogs that investigate V2X communication and learn what memory is helping to make it possible.

About Our Blogger

James Hawley

James is an Automotive Strategic Marketing Manager for Micron's Embedded Business Unit.

Login or Sign Up Now for an account to leave a comment.