Opportunities and challenges
in smarter connected control
By Giles Peckham,
Xilinx
Smarter connected control is
needed to satisfy the increas-
ing commercial and consumer
demands of manufacturing,
transport and communica-
tions. Xilinx is addressing this
challenge with the Zynq-7000
All Programmable SoC, the
first device integrating an
ARM dual-core Cortex-A9
MPCore processor with
programmable logic and key
peripherals on a single chip.
February 2014
18
M
ICROCONTROLLERS
The desire for more and better is intrinsic to
human nature, and is a powerful driver de-
manding more feature-rich, powerful and cost-
effective systems within numerous aspects of
life. Broadcast technologies, for example, are
required to deliver increasingly lifelike and en-
joyable viewing experiences. Communications
systems must support richer and more imme-
diate connections between people. Cars and
public transport are expected to offer greater
safety and enjoyment with better environmen-
tal credentials. Manufacturing technology needs
to deliver continuous improvements against
metrics such as throughput and quality control.
And our desire for unfailing security whether
at home, at work, or in public places demands
all-seeing detection with the ability to record
events and recognise or ideally identify of-
fenders.
Smarter systems combining high-speed com-
puting, local decision-making capability and
high-bandwidth network connectivity hold
the key to meeting many of these increasing
demands. Communications systems of all types,
from consumer broadband services to net-
worked manufacturing automation need to
recognise, process and route diverse types of
data traffic efficiently, with awareness of any
bottlenecks in the network. As another example,
precision positioning and motion controls in
industrial robotic equipment and automated
machinery require high-performance control
systems capable of rapidly executing digital
motor-control algorithms in multiple axes si-
multaneously. On the other hand, powerful
image-recognition algorithms – often operating
in real-time – are critical for equipment such
as advanced driver assistance systems in the
latest cars, industrial automation and quality
control systems, and networked video surveil-
lance cameras.
The increasing expectations placed on smarter
connected applications across numerous in-
dustrial markets are driving demands for high-
speed processing such as digital imaging and
analytics, as well as essential control processing,
to be performed closer to the network edge,
rather than following a conventional centralised
architecture. Some examples include networked
security surveillance systems capable of per-
forming basic processing on full-colour, high-
resolution video to detect suspicious activity,
and forward only the relevant data to a central
controller performing higher-level functions
such as comparing data from the captured im-
ages against a biometric database. Another is
pedestrian recognition in driver-assistance sys-
tems, which must apply advanced high-speed
algorithms to distinguish pedestrians from
other objects such as road furniture or parked
vehicles. In the manufacturing automation
field, equipment such as assembly equipment
or food-preparation systems use sophisticated
image-recognition algorithms to inspect man-
ufactured items at high speed, or sort items
such as harvested apples by size, attractiveness
or presence of any defects. Safety systems also
are adopting smart connected technologies to
allow greater flexibility and improve protection
for factory workers. Examples include virtual
safety barriers implemented using one or more
video cameras to monitor the space around
machines in a production area, or other hazards
that cannot be satisfactorily protected using
physical barriers. A virtual system can allow
unrestricted movement of robotic equipment,
for example, yet is able to issue a warning and
ultimately shut down the machine if the bound-
ary is breached by a human operator.
In cases such as these, real-time response is
typically needed but is increasingly difficult to
achieve using a conventional centralised and
software-based approach to image processing.
For some time now, designers of high-perfor-
mance embedded systems – particularly real-
time systems – have been using FPGAs to ac-
celerate functions that cannot be performed
quickly enough in the main processor or DSP.
It is also worth noting that safety-critical sys-
tems are significantly easier to validate when
implemented in hardware, allowing developers
to avoid the need for rigorous testing of safe-
ty-critical software.
