Forward facing Advanced Driver Assistance Systems (FF ADAS) are
becoming increasingly important, especially as the European New Car
Assessment Programme (Euro NCAP) ratings are tied to AEB (Autonomous
Emergency Braking) and other important car safety features. Euro
NCAP recently announced a new rating scheme for the years 2013 to
2017, which is also driving ADAS trends for players across the
automotive market, from tier 1s and manufacturers, to semiconductor
ADAS applications such as AEB and Lane Departure Warning (LDW) will
be taken much more strongly into account by 2014, and will be key
for achieving a top Euro NCAP 5-star rating, which significantly
impacts car manufacturers’ sales rates. Along with AEB and LDW,
other features such as Intelligent Highbeam Control (IHC) and Road
Sign Detection (RSD) are now integrated on most standard FF ADAS
cameras. Low- to mid-range systems of the near future will possess
at least some version of AEB (described below), LDW and IHC.
It is evident that the auto industry is under pressure to find ways
to implement these highly advanced, computationally and network
intensive systems without driving up the bill-of-materials (BOM) per
vehicle. Indeed, the introduction of these new applications
will likely evolve as an overall cost reduction in the ADAS camera
platform. Ideally, automakers will want to have a common ‘smart’
platform from which they can introduce and deploy systems
differentiated by feature sets across their various vehicle classes.
This means a single platform design for a smart FF ADAS processing
camera that is low power, has low BOM cost, and is scalable across
vehicle classes, especially low to mid-range.
There is a host of design challenges for a FF ADAS processing camera
targeted at the applications described above. System design
requirements which shape the final solution include:
High-definition (HD) imaging and multi-function support:
Applications like people detection, and especially road sign
detection, drive higher resolution image sensor requirements.
Existing implementations are VGA but are rapidly migrating to
megapixel and beyond in next generation systems. In
addition with multiple functions, there is a requirement for the
system to scale and partition image frames real time in advance
of algorithmic processing. Along with increasing image sensor
resolution, expect higher frame rates (for example, up to 60fps)
to support functions like pedestrian detection at higher
vehicular speeds (for example, 60km/h for NCAP testing).
Computational support: Vision processing requires a careful
mapping of software to the embedded hardware to meet demand for
very large computational requirements.
Data flow: Depending on the implementation, there is likely a
requirement to support high bandwidth data movement especially
between sensor interface, system memory and processor core.
Differentiation: Tier-1’s and OEM’S must differentiate and run
non-standard analytics algorithms. This requires open,
programmable systems that support easy porting of their in-house
algorithm code base.
Temperature sensitivity: Smart camera systems are
traditionally situated against the windshield in front of the
rear view mirror. This is one of a vehicle’s more
challenging thermal environments, due to direct sun and
environmental exposure. Systems, therefore, must be temperature
Power sensitivity: The thermal requirements mentioned above
necessitate a tight control of power dissipation. This is
particularly true given the sensitivity of higher resolution
image sensors to thermal noise.
Low cost: Cost is a major factor in many markets. Due to the
cost-sensitive nature of FF ADAS, manufacturers expect silicon
optimized to provide maximum features at appropriate price
ranges and within compact chip form factors. Devices with
cores that are specific to programmable vision processing at low
power and low die area are typically more competitive.
Functional safety: Vision-based ADAS applications are
moving beyond providing information to the driver through useful
LEDs to life-critical active safety systems. FF ADAS
processing cameras will be subject to the same safety
considerations that have existed with radar systems (for
example, for ACC).
From one hand, this driver-assistance will help in different situations. From the other hand, year by year a driver is less and less capable to react on the road. Machines take leading positions, soft rules the speed and braking. I'm afraid In 10 years all cars will be autonomous. I say it's dangerous! travel insurance
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.