Graphics Reference
In-Depth Information
end, optimization for the target processor architecture will significantly improve the
system performance.
The algorithmdevelopment represents only a small fraction in ADAS design cycle
as the major challenge in the development of these systems is to take into account the
variety of traffic scenarios. Extensive test and validation is an enormous undertaking
and the most challenging aspect of ADAS development, especially when it comes to
the vision systems. In an effort to test all scenarios and to achieve 100% accuracy and
zero false positives, under all possible conditions, thousands of hours of video clips
must be gathered and run in regression test database. As the test databases would not
be able to cover all test cases, suppliers spend years of testing and validating systems
and performing real-world field trials.
3.5 Conclusion
We have come a longway since first demonstration of an autonomous car at NewYork
World's Fair in 1939. Newkinds of sensors and state-of-the-art embedded processors,
with sufficient compute power to support rapid growth of computer vision, are factors
catalyzing growth of camera-based advanced drivers systems.
Regulationwill further drive the proliferation of multifunction cameras to themost
cost-sensitive segment of the automotive market, which is also the most crash vul-
nerable segment. The propagation is led by New Car Assessment Program (NCAP)
changes inEurope, but introduction of cameras in themainstreamvehicles is expected
to become noticeable in the USA too, driven by the National Highway Transportation
Safety Administration (NHTSA), the National Transportation Safety Board (NTSB),
and the Insurance Institute for Highway Safety (IIHS).
In order to reduce the number of backover injuries and deaths and to improve
visibility when a car is backing out of a driveway or parking space, NHTSA issued a
rule which mandates use of rear view cameras in all new vehicles in the US starting
fromMay 2018. The regulation, depending on final performance requirements, might
stimulate spread of passive or active rearview camera systems.
Projects such as Prometheus, ARGO [
51
], and DARPA Urban Challenge, were
instrumental in helping Google's driverless vehicle come to life. Today, many major
automotive manufacturers including Mercedes-Benz, BMW, Audi, Volkswagen,
Ford, GM, and Toyota are testing driverless cars.
Eventually, wemight see a circuit fail and that one defect in one billion could cause
a fatal crash. Who is responsible when things go wrong? How should the cars with
such systems be tested? The number of electronic components networked together
in vehicles together with communication systems between cars, creates opportuni-
ties that could potentially be exploited by a malicious attacker. Cyber security is a
concern and several researchers have shown possible ways to compromise security
of automobiles and hijack functions such as steering or braking by exploiting a broad
range of attack vectors, including CD players, Bluetooth, and cellular radio [
52
,
53
].
