23
I
ndustrial
I
o
T
April 18
available to receive and process control packets when they arrive on
a TSN-enabled port. The ability to respond to control packets also
helps the CPU to address events coming to the processor from other
inputs and to execute loops controlling the system the processor is
part of. These loops may need to run up to every 30 microseconds
or faster - a degree of precision that a conventional IT-derived oper-
ating system cannot meet. The need for more automation requires
increased processing capabilities in embedded controllers. Higher
performance processing can be used to reduce control loop timing,
moving robotic arms and assembly lines faster and increasing factory
output. It can also increase the number of axes managed by a sin-
gle motion controller, leading to robots with more articulated joints,
which can operate in tighter spaces or perform tasks that the previous
generation of factory robots could not address. Robots that can learn
tasks from a human operator will require image processing, along
with new machine learning algorithms.
Commercial RTOSs include VxWorks from Wind River and Nucleus
from Mentor Graphics. These vendors have a long history of sup-
porting the NXP QorIQ family and its predecessors. With the emer-
gence of industrial-grade Linux, open-source alternatives are another
option. These provide industrial enterprises and OEMs the agility to
add new capabilities to their systems. Unlike IT-focused and non-re-
al-time embedded Linux distributions, industrial-grade ones provide
the determinism, manageability, industrial networking, and security
required of OT.
One approach to adding real-time capability to Linux is to apply the
PREEMPT_RT patch to the kernel to eliminate situations where a
software process is blocked indefinitely by another process. In this
scheme, applications are coded to the usual Linux API. Another
approach taken by Xenomai is to add classic RTOS APIs to a Linux
system, facilitating porting traditional RTOS applications to Linux.
This supplier also provides mechanisms for device drivers to respond
to peripherals in real time, firming up the real-time guarantees Linux
can offer. To ease the transition to Linux from a classic RTOS, NXP is
working with the industrial Linux community on a distribution inte-
grating the various real-time enhancements and TSN stacks while
maintaining standard Linux capabilities.
Processing capacity must also be available for analytics. The IoT is
not only about networking embedded systems but about captur-
ing data from sensors, analyzing the data, and directing the system
responses. A common notion is that distant servers in the cloud per-
form the analysis. However, the amount of data to be transported
and analyzed, the time-criticality of the decisions to be made, and
the proprietary nature of the data will lead manufacturers to pro-
cess manufacturing data locally. Analysis could be done not only on
a computer at a factory site but even within production machinery,
given sufficiently powerful processors. Beyond analysis, processing
capacity in an Industry 4.0 regime will be used to manage operations
remotely, to enable machines to coordinate among themselves auton-
omously, and to gain efficiencies from linking production data and IT
systems such as those for enterprise resource planning.
Another function demanding processing power is the human-ma-
chine interface (HMI). Smartphone-inspired interfaces will increas-
ingly permeate the staid world of industrial equipment. Easy-to-use,
visual interfaces simplify operator control of machines. High-resolu-
tion screens enable viewing the output of high-definition (or better)
cameras inspecting goods as they are manufactured. Driving these
screens will be the same type of graphics processing units (GPUs)
found in smartphones. Although this 3D performance of GPUs will
be scaled down from what is in a smartphone to reduce cost and
power, they will support large, high-resolution screens; overlays of
