Accelerating Performance with the Artificial Intelligence of Things
Onboard analytics, in some cases, to
train and run AI models at the edge. 6 7 transmit their experiences to other cars in
the network.
The physical components are amplified by
the smart elements. The smart elements are
in turn amplified by connectivity, which
enables
monitoring,
control
and
optimization. But by itself, connecting things
does not promote learning. It paves the way,
but that is just the foundation. These capabilities are the basis for the
personalization required of IoT applications:
As humans, we want to be treated
individually and know that our habits,
behavioral patterns and preferences are
considered. For instance, think about a
consumer wearable technology that
monitors movements to detect signals of an
impending injury in an athlete. No two
humans move the same, so the application
would only be meaningful with great
personalization. For another example,
retailers use IoT-enabled cameras for object
detection along with machine learning to
deliver tailored advertisements and offers to
shoppers at the right moment.
At the most basic level, the data generated
from IoT devices is used to trigger simple
alerts. For example, if a sensor detects an
out-of-threshold condition, such as
excessive heat or vibration, it triggers an
alert and a technician checks it out. In a more
sophisticated IoT system, you might have
dozens of sensors monitoring many aspects
of operation.
As machines become more complex, they
need personalization too. Two pieces of
industrial equipment of the same make and
models do not perform identically under
different conditions and might not be used
in the same way. Treating them alike misses
IoT opportunities for greater operational
efficiency, enhanced safety and better use of
resources. For example, when producing
semiconductor wafers, AIoT improves yield
by determining the optimal path for wafer
lots to travel during the manufacturing
process. This eliminates scrap waste and
optimizes product quality.
All these scenarios add value to and from
connected devices. But the real value of IoT
comes at yet another level of sophistication.
It happens when devices learn from their
specific use or from each other and then
automate actions. It happens when they can
adapt, change behavior over time, make
decisions, act and tune their responses
based on what they learn.
For example, a model using IoT data to
detect failures can push machine controls to
the appropriate IoT powered actuators to
reduce the possibility of failures on similar
equipment. Self-driving vehicles can
6
IIC, Edge Computing Task Group https://www.iiconsortium.org/pdf/Introduction_to_Edge_Computing_in_IIoT_2018-06-18.pdf
7
Aslett, Matt. 451 Research https://www.sas.com/content/dam/SAS/documents/marketing-whitepapers-ebooks/third-party-
whitepapers/en/data-analytics-at-the-edge-110405.pdf
- 35 -
June 2019