Graphics Reference
In-Depth Information
sensors operating at different resolutions, frame rates, under different lighting con-
ditions, and with different optics. Embedded vision algorithms are the “brains” of
ADAS along with components such as software framework, real-time operating sys-
tem, and tools.
The development of advanced driver assistance system starts with a definition of
the requirements outlining functions and operational constraints. Hazard and risk
analysis is then performed in order to determine the system safety requirements.
These key components of the vision-based driver assistance systems are discussed
in this section.
3.4.1 Functional Safety
Advanced driver assistance systems can take over elements of driving such as steering
or braking where undetected dangerous failure can be catastrophic. When designing
a vision system for ADAS, how can we provide the evidence that risks have been
minimized? Functional safety has immense implications on all components in ADAS
and safety must be embedded in the culture of every organization and the supply
chain. For example, offset error of just one pixel when calculating disparity could
generate a wrong assessment of distance between the car and a pedestrian, resulting
in a disaster.
The development of ADAS systems is governed by international safety standard
for road vehicles ISO 26262. Implementing ISO 26262 allows leveraging a com-
mon standard to measure how safe a system will be in service. The ISO 26262
standard provides regulations and recommendations throughout the product devel-
opment process from conceptual development through decommissioning. It details
how to assign an acceptable risk level to a system or component and document the
overall testing process. ISO 26262 also contains detailed guidance on software tool
qualification. The objective of tool qualification is to provide evidence that a software
tool is suitable for use in the development of safety-related software according to the
standard.
In order to classify risk level and to quantify the degree of rigor that should be
applied in development, implementation, and verification, the ISO 26262 defines
Automotive Safety Integrity Levels (ASIL). There are four ASIL levels identified by
the standard: A, B, C, and D. ASIL A dictates the lowest integrity requirements on
the product while ASIL D dictates the highest safety requirements. The ASIL level is
established by performing a risk analysis of a potential hazard by looking at severity,
exposure, and controllability of the vehicle operating scenario.
ASILs can be decomposed over a system and high ASILs can be met by multiple
redundant components working together, each with lower ASIL.
