SAE Update - April 2022 - 23

Embedded System Design Challenges
in the age of Open-Source and AI
TECH FOCUS:
SOFTWARE
massive complexity of the components. The problems
become even more difficult when viewed from a
product lifecycle point of view because the SW
paradigm introduces constant additions to the supply
chain and release points. Managing this complexity is
a focus for most embedded design teams.
Ai paradigm
Recent advancements in machine learning (a key
subdomain of AI) have enabled the harnessing of
sensors to build intelligence on the edge. With these
capabilities, sound can be interpreted as language
(NLP), video can lead to object recognition (vision),
and there are many more applications.
This AI/ML paradigm is built on top of another
abstraction above the processor called the AI model
(CNN, Transformer, etc). There is a compile path
(training) with a run-time flow (inference). For AI, the
data is used to produce the algorithm, so there is a
requirement to manage the training data in a similar
fashion as software code in the past.
From a system design perspective, AI/ML is at the
center of most of the interesting applications in
embedded systems, with key focus points around
autonomy, cybersecurity, and safety systems. The AI
Phase adds another independent ecosystem onto the
system designer plate which is independent from the
HW and SW phases and operates at its own cadence.
The data/training cycle updates can range from being
tied to SW release cycles to near continuous
operation. Today, traditional EDA systems do not
consider the AI paradigm.
where does this leave the embedded
design space ?
The traditional HW paradigm is well understood by
UPDATE
the embedded design
community, and there are
strong enabling tools from the
EDA and PLM communities. As
OEMs such as Tesla have shown,
a highly leveraged SW paradigm
can be very powerful to enable
both functionality and
innovation. However, the critical
enabling tools from EDA and
PLM struggle with managing
the SW paradigm. Specifically,
the fundamental processes for
design discovery, mapping of
function to component,
component selection, and
lifecycle management of the
product must be upgraded.
Finally, the AI paradigm is at
the forefront of embedded
system design and, largely
speaking, there is no real
capability available from the
EDA and PLM sectors to enable
these capabilities.
Overall, the SW and AI
paradigms have the promise to
enable the third mega-wave of
capability for the benefit of the
world. However, these
fundamental shifts in system
design methodology must have
the critical enabling
technologies in EDA and PLM to
deliver on the promise of these
technologies. n
April 2022
23
SAE Update - April 2022

Table of Contents for the Digital Edition of SAE Update - April 2022
