“Multimedia will become more predominant
in cars and Intel will support
this with a low-power open platform that
seamlessly connects all types of AI products
from different vendors,” says Staci
Palmer, director of in-vehicle infotainment
for Intel’s low-power products division.
“A platform architecture is needed that
can accommodate continuing electronic
innovations and is compatible with longer
automobile design cycles.”
The Atom processor is part of Intel’s
strategy of making available low-power
ICs for such platforms. Based on the Intel
Architecture (IA) and fabricated using
45-nm design rules, the CMOS processor
can operate at up to 2 GHz at 1 V, yet
it dissipates very little power (from 0.6 to
2 W) (Fig. 4). Intel and Wind River Systems
are collaborating on an open-source
Linux platform for AI systems. The BMW
Group, Bosch GmbH, Delphi, and Magneti
Marcelli actively support Linux for AI
applications, too.
Altera selected the Media Local Bus
(MediaLB) IP core from Standard Microsystems
(SMSC) for its PARIS AI development
platform. The MediaLB interchip
communication technology efficiently
transports multimedia data through
SMSC’s intelligent network-interface
controllers (INICs) and onto SMSC’s
Media Oriented Systems Transport
(MOST) network.
Furthermore, SMSC’s INIC eLITE
technology eliminates extra wiring and
the added cost of transmitters, receivers,
analog-to-digital converters (ADCs), and
digital-to-analog converters (DACs) in
networks like MOST. It makes it easier
for a designer to add additional nodes on a
network with minimal cost.
SOFTWARE AND OTHER CHALLENGES
As AI system hardware increases, software
takes on a bigger role. “We’re seeing 32-bit
microprocessors with memory management
units for networking in AI systems,
many running under Unix,” says Andrew
Poliak, worldwide automotive sales director
for QNX Software. He also notes that
tier-one suppliers and OEMs are looking
for more middleware products to enhance
their offerings more cost-effectively and to
improve the user-AI voice interface.
That’s the reason why QNX released its
Aviage acoustic processing kit, which the
company says reduces the cost and improves
the quality of hands-free systems. “Removing
voice and echo cancellation in voice
communication is a trend in AI systems, and
enhancements in this area will continue,”
explains Poliak. “As you make voice communication
clearer and more intelligible
to the driver, it decreases the driver’s inattention
and distraction and thus increases
driving safety.”
One challenge is creating a safe environment
for drivers who want to download
entertainment and business information
from the Internet and into the car via a
mobile phone. Future AI systems will need
more intelligence as the automobile interacts
further with the Internet, gains communications
capabilities, and automatically
executes more functions. For example,
navigation systems will provide up-to-date
traffic and roadway conditions, in addition
to assessing the driving route and suggesting
alternatives.
Software-defined radio (SDR), which
has been around for many years, will play
a larger role in AI systems. SDR will
enable designers to reconfigure the same
hardware for different communications
protocols, standards, and applications via
innovative RF algorithms implemented
in CMOS technology. The True Software
Radio technology from Terocelo Inc. targets
such applications.
As part of the company’s Lycon line,
these chips are designed to dramatically
improve the way wireless signals are transmitted
and received. True SDR chips make
it possible for software commands to fully
control and reconfigure wireless transmitters
and receivers, as well as radio signal
processing. They replace the analog front
end, intermediate frequency (IF) processing,
analog-to-digital conversion, and digital
filtering sections of today’s conventional
radio transmitters and receivers.