Edge Intelligence: The Central Cloud is Dead – Long Live the Edge Cloud!
(3) Open platform for edge computing
ECNs are machines deployed close to the
end device of an infrastructure, having the
role to take local decisions based on sensed
or received information and policies or
algorithms from the core servers. The
decisions can be applied (physically) locally
or communicated to the core servers for
aggregation and final decision.
At present, there are multiple vendors of
edge computing products, and the edge
computing
(program
execution)
environment of those products is
proprietary. This situation makes it difficult
for developers of edge computing
applications to develop a portable
application which can run across multiple
edge computing products. It would be
desirable to utilize the wisdom of people
from various disciplines in system
development by enlarging the developer
community, as in the case with Smartphone
applications. Making the edge computing
environment open can contribute to that
purpose. Additionally, measurement such as
certification of an application to guarantee
that the application does not behave
maliciously and is not harmful to a system,
and lifecycle management of an application
would need to be in place.
(2) Evolutionary lighting policies and energy
management policies
The basic lighting policy is drawn up based
on the time, date and geographical location.
Some reactive policies can be made
according to the environmental brightness
and the proximity of vehicles and
pedestrians. The edge and the cloud should
have learning abilities to build continuously
evolving rules. Since lamps and the other
modules, especially the charging module,
are part of the smart grid, the energy
management policies should also be
optimized based upon predicted use.
Smart City Lighting
(3) Cloud offloading and privacy: process the
data locally at the edge
(1) Function add and removal for Pole
networking
The allocation of processes between the
edge and the cloud must be defined: the
time-critical data must be processed at the
edge to get a timely response; some lower
stage processing can be conducted at the
edge, such as filtering and aggregation, so
that the cloud can be offloaded; for privacy
reasons, some data must be processed
locally, e.g. in video surveillance, only
abnormal events are reported to the cloud
while the citizens’ portrait should be
protected.
When deploying lighting poles, the installed
function modules should first be identified,
and then neighboring devices and Edge
Compute Nodes (ECNs) specified with which
to establish connections. In this way, the
network can be built without much human
intervention. The networking topology
respond to conditions such as connection
qualities
and
the
computation
load/capabilities of ECNs. Once a function
module is added or removed from a pole,
the network and the applications at the ECN
should change accordingly.
- 26 -
September 2017