Edge Intelligence: The Central Cloud is Dead – Long Live the Edge Cloud!
traditional manual inspection is no longer
necessary;
therefore
operational
expenditures can be greatly reduced.
Thirdly, if connectivity to the lights is lost,
the systems can continue operating by using
the policy from the edge device, and not rely
upon the cloud or datacenter.
by artificial intelligence at the cloud and then
downloaded by the edge computing node.
Indoor Location Tracking
The bandwidth requirements for indoor
location
tracking
are
moderate:
approximately 2 MB, with very low latency
(<1 ms) and low contention. The system
requires a backhaul of trilateration data for
a number of sensor sources (all normalized
to IP/UDP packets) and conversion into a
high quality location estimate. For high value
asset
tracking,
real-time
location
computation require mathematical results
and do not afford the delay introduced by
communication to the cloud. As a result,
having a gateway which processes the
sensor samples as close as possible to the
source, while maintaining the connection
with the cloud, or at least outside the
customer premises, are critical for a system
of this type. Within the gateway module is an
embedded, tunable, machine intelligence
module to perform the location estimation,
which then forwards real world positions
and user status to the administrative/UI
module in the cloud. The model may also be
tuned and the Machine Intelligence (MI)
module updated.
In the future, light poles will move beyond
single functions (the lighting). Many other
functional modules can be added, such as
environment/utility monitoring,
video
surveillance, vehicle-to-infrastructure (V2I)
communication devices, and so on, making
the light poles become an integrated system
of sensing and service provision.
Smart Elevators
Taller buildings and access make elevators
indispensable in cities. The operation and
maintenance of elevators is considerably
expensive due to manual inspection, fault
detection and repairs. Smart elevator with
edge and cloud intelligence allow vendors to
upgrade from inefficient, expensive
preventive maintenance models to next-
generation, real-time, targeted, predictive
maintenance, extending value from
products to services.
Hundreds of sensors are deployed to
monitor the elevator’s status. Based on this
data, the edge computing node is capable of
detecting potential device faults early and
sending out the alarms immediately. When
the edge computing node fails to connect to
the cloud, the data can be stored locally until
the connection recovers. By analyzing the
historical data at the cloud, faults can even
be predicted, so that maintenance is given
accordingly before a fault actually occurs.
New features of faults can also be extracted
Lone Worker Safety
For lone worker safety an intelligent
gateway is used to receive signals indicating
the location of a particular employee. In such
cases an MI module would also be
embedded in the GPS signal transmitter,
which would use an MI module to
characterize the wearer’s gait and
orientation. The module would “learn” over
a period the “normal” behavior of an
individual, and thus be able to generate an
- 24 -
September 2017