LONDON -- As vehicles become highly independent and begin to drive and react to traffic on their own, autonomous systems will aggregate and process data from a variety of on-board sensors and connected infrastructure. This forces the industry to hit a hard reset on advanced driver assistance systems (ADAS) architectures, currently dominated by distributed processing and smart sensors. Automotive OEMs will need to adopt new platforms based on powerful, centralized processors and high-speed low latency networking. ABI Research forecasts 13 million vehicles with centralized ADAS platforms will ship in 2025.
"The distributed approach to ADAS systems will prove unsustainable as OEMs look to deliver highly automated driving around 2020," says James Hodgson, Industry Analyst at ABI Research. "The new centralized ADAS architectures will unify sensing, processing, and actuation to deliver integrated decision-making for smooth path planning and effective collision avoidance."
This transition will present major opportunities for vendors new to the industry, as well as old incumbents, including NVIDIA, NXP, and Mobileye, who all announced centralized autonomous driving platforms. While each is in a different stage of development, all have common themes emerging, particularly in relation to processing power. The platforms average between eight and twelve teraflops (TFLOPs)—a figure that is orders of magnitude beyond the typical smart sensor currently deployed in ADAS.
Physical separation of numerous dumb sensors and centralized processing will also open up opportunities for in-vehicle networking vendors. Ethernet-based solutions from vendors such as Marvell Semiconductor and Valens Semiconductor are well-positioned to meet the needs of high bandwidth and stringent automotive-grade requirements at a low cost.
"We are fast approaching the end of what can be achieved in automation within the confines of legacy architectures," concludes Hodgson. "While there are not yet any specific standards for centralized ADAS, it is interesting that three separate Tier 2s announced very similar platforms in quick succession. Vendors across the ecosystem need to take this time to plan accordingly in order to appropriately manage the industry transition toward centralized ADAS architectures."
These findings are from ABI Research's ADAS Vehicle Architectures—Smart Sensors versus Centralized Platforms.