Edge Intelligence: The Central Cloud is Dead – Long Live the Edge Cloud!
away from the cloud to the edge of the
network, continues to expand.
I NTRODUCTION
Edge intelligence allows bringing data (pre-)
processing and decision-making closer to the
data source, which reduces delays in
communication. In addition, such (pre-)
processing makes it possible to accumulate
and condense data before forwarding it to
Internet of Things (IoT) core services in the
cloud or storing it, which perfectly matches
the capacities offered by the upcoming fifth
generation wireless technology (5G)
networks providing localized throughput
and delay enhancements. Edge computing
makes processing and storage resources
available in close proximity to edge devices
or sensors, complementing centralized cloud
nodes and thus allowing for analytics and
information generation close to the origin
and consumption of data. Supplementary
resources may even reside on end devices
that might not be continuously connected to
the backbone network. Additionally, edge
intelligence allows future applications to
depend on context awareness capabilities
for mutual detection and proximity services,
(near) real-time responsiveness for a tactile
Internet, data analytics at the edge and/or
end
device
and
device-to-device
communication capabilities.
T REND D RIVERS AND S TATE OF THE
A RT FOR E DGE I NTELLIGENCE
This article analyses manufacturing, smart
building, asset management, smart grid and
transportation industrial verticals. Some of
the most stringent needs of the analyzed
industries that can be solved through placing
intelligence on the edge computing units
include:
As processors, microcontrollers and
connectivity modules are embedded into a
plethora of new devices, the application of
edge intelligence in smart appliances,
wearables, industrial machines, automotive
driver assistance systems, smart buildings
and the like continues to increase. To enable
and realize IoT’s true value, the trend toward
adopting edge intelligence, which pushes
processing for data-intensive applications
- 18 -
Mobility: industries demand in terms of
networking support (mobility and
wireless broadband) is a high degree of
mobility, especially in terms of
handovers. At the same time, the
“quality of service” and session handover
management are critical aspects that can
benefit from intelligence in the network
components.
Ultra-low latency in decision-making:
decisions on detection or actuation have
to be taken within a delay of less than
tens of milliseconds. For this, intelligence
residing at the edge can help lower the
delay and achieve the targeted response
time.
Autonomy: a key requirement for use
cases is autonomous to continuing
operation without connection to core
server or service, to prevent damage to
persons, goods or infrastructure.
Security: security is a feature that can
never receive too much attention.
Access control to physical or virtual
resources (e.g. data) has to be ensured.
Locally provisioned or learned policies
and
other
mechanisms
for
September 2017