Scientific Reports,
Journal Year:
2025,
Volume and Issue:
15(1)
Published: May 1, 2025
Abstract
Mobile
Edge
Computing
(MEC)
systems
face
critical
challenges
in
optimizing
computation
offloading
decisions
while
maintaining
quality
of
experience
(QoE)
and
energy
efficiency,
particularly
dynamic
multi-user
environments.
This
paper
introduces
a
novel
Adaptive
AI-enhanced
(AAEO)
framework
that
uniquely
integrates
three
complementary
AI
approaches:
deep
reinforcement
learning
for
real-time
decision-making,
evolutionary
algorithms
global
optimization,
federated
distributed
knowledge
sharing.
The
key
innovation
lies
our
hybrid
architecture’s
ability
to
dynamically
adjust
strategies
based
on
network
conditions,
user
mobility
patterns,
application
requirements,
addressing
limitations
existing
single-algorithm
solutions.
Through
extensive
MATLAB
simulations
with
50–200
mobile
users
4–10
edge
servers,
demonstrates
superior
performance
compared
state-of-the-art
methods.
AAEO
achieves
up
35%
improvement
QoE
40%
reduction
consumption,
stable
task
completion
times
only
12%
increase
under
maximum
load.
system’s
security
analysis
yields
98%
threat
detection
rate,
response
100
ms.
Meanwhile,
reliability
metrics
indicate
99.8%
rate
mean
time
failure
1,200
h.
These
results
validate
the
proposed
approach’s
effectiveness
complex
next-generation
MEC
systems,
heterogeneous
environments
requiring
adaptation.
IEEE Access,
Journal Year:
2023,
Volume and Issue:
11, P. 12555 - 12586
Published: Jan. 1, 2023
Fog
computing
has
emerged
as
a
paradigm
for
resource-restricted
Internet
of
things
(IoT)
devices
to
support
time-sensitive
and
computationally
intensive
applications.
Offloading
can
be
utilized
transfer
resource-intensive
tasks
from
resource-limited
end
resource-rich
fog
or
cloud
layer
reduce
end-to-end
latency
enhance
the
performance
system.
However,
this
advantage
is
still
challenging
achieve
in
systems
with
high
request
rate
because
it
leads
long
queues
nodes
reveals
inefficiencies
terms
delays.
In
regard,
reinforcement
learning
(RL)
well-known
method
addressing
such
decision-making
issues.
large-scale
wireless
networks,
both
action
state
spaces
are
complex
extremely
extensive.
Consequently,
techniques
may
not
able
identify
an
efficient
strategy
within
acceptable
time
frame.
Hence,
deep
(DRL)
was
developed
integrate
RL
(DL)
address
problem.
This
paper
presents
systematic
analysis
using
DRL
algorithms
offloading-related
issues
computing.
First,
taxonomy
offloading
mechanisms
based
on
divided
into
three
major
categories:
value-based,
policy-based,
hybrid-based
algorithms.
These
categories
were
then
compared
important
features,
including
problem
formulation,
techniques,
metrics,
evaluation
tools,
case
studies,
their
strengths
drawbacks,
directions,
mode,
SDN-based
architecture,
decisions.
Finally,
future
research
directions
open
discussed
thoroughly.
Alexandria Engineering Journal,
Journal Year:
2024,
Volume and Issue:
91, P. 12 - 29
Published: Feb. 6, 2024
Energy
efficiency
is
a
key
area
of
research
aimed
at
achieving
sustainable
and
environmentally
friendly
networks.
With
the
rise
in
data
traffic
network
congestion,
IoT
devices
with
limited
computational
power
energy
resources
face
challenges
analyzing,
processing,
storing
data.
To
address
this
issue,
computing
technology
has
emerged
as
an
effective
means
conserving
for
by
providing
high-performance
capabilities
efficient
storage
to
support
collection
processing.
As
such,
energy-efficient
computing,
or
"green
computing,"
become
focal
point
researchers
seeking
deploy
large-scale
This
study
provides
comprehensive
Survey
recent
efforts
green
best
our
knowledge,
none
studies
literature
have
discussed
all
types
(edge,
fog,
cloud)
their
role
enabling
massive
networks
terms
efficiency.
The
article
starts
overview
technologies
then
goes
discussion
empowering
energy-saving
techniques
environments
including,
energy-aware
architecture,
aggregation
compression,
low-power
hardware,
scheduling,
task
offloading,
switching
on/off
unused
resources,
virtualization,
harvesting,
cooling
optimization.
outline
roadmap
toward
realizing
vision
environment
networks;
addition,
open
door
interested
follow
continue
Energy-Efficient
Computing.
Applied Sciences,
Journal Year:
2022,
Volume and Issue:
12(18), P. 9124 - 9124
Published: Sept. 11, 2022
New
technologies
bring
opportunities
to
deploy
AI
and
machine
learning
the
edge
of
network,
allowing
devices
train
simple
models
that
can
then
be
deployed
in
practice.
Federated
(FL)
is
a
distributed
technique
create
global
model
by
from
multiple
decentralized
clients.
Although
FL
methods
offer
several
advantages,
including
scalability
data
privacy,
they
also
introduce
some
risks
drawbacks
terms
computational
complexity
case
heterogeneous
devices.
Internet
Things
(IoT)
may
have
limited
computing
resources,
poorer
connection
quality,
or
use
different
operating
systems.
This
paper
provides
an
overview
used
with
focus
on
resources.
presents
frameworks
are
currently
popular
provide
communication
between
clients
servers.
In
this
context,
various
topics
described,
which
include
contributions
trends
literature.
includes
basic
designs
system
architecture,
possibilities
application
practice,
privacy
security,
resource
management.
Challenges
related
requirements
such
as
hardware
heterogeneity,
overload
resources
discussed.
IEEE Open Journal of the Communications Society,
Journal Year:
2023,
Volume and Issue:
4, P. 2609 - 2666
Published: Jan. 1, 2023
Technology
solutions
must
effectively
balance
economic
growth,
social
equity,
and
environmental
integrity
to
achieve
a
sustainable
society.
Notably,
although
the
Internet
of
Things
(IoT)
paradigm
constitutes
key
sustainability
enabler,
critical
issues
such
as
increasing
maintenance
operations,
energy
consumption,
manufacturing/disposal
IoT
devices
have
long-term
negative
economic,
societal,
impacts
be
efficiently
addressed.
This
calls
for
self-sustainable
ecosystems
requiring
minimal
external
resources
intervention,
utilizing
renewable
sources,
recycling
materials
whenever
possible,
thus
encompassing
sustainability.
In
this
work,
we
focus
on
energy-sustainable
during
operation
phase,
our
discussions
sometimes
extend
other
aspects
lifecycle
phases.
Specifically,
provide
fresh
look
at
identify
provision,
transfer,
efficiency
three
main
energy-related
processes
whose
harmonious
coexistence
pushes
toward
realizing
systems.
Their
related
technologies,
recent
advances,
challenges,
research
directions
are
also
discussed.
Moreover,
overview
relevant
performance
metrics
assess
energy-sustainability
potential
certain
technique,
technology,
device,
or
network,
together
with
target
values
next
generation
wireless
systems,
discuss
protocol,
integration,
implementation
issues.
Overall,
paper
offers
insights
that
valuable
advancing
goals
present
future
generations.
IEEE Transactions on Vehicular Technology,
Journal Year:
2023,
Volume and Issue:
72(8), P. 10696 - 10709
Published: March 8, 2023
To
handle
the
ever-increasing
IoT
devices
with
computation-intensive
and
delay-critical
applications,
it
is
imperative
to
leverage
collaborative
potential
of
edge
cloud
computing.
In
this
paper,
we
investigate
dynamic
offloading
packets
finite
block
length
(FBL)
in
an
edge-cloud
collaboration
system
consisting
multi-mobile
(MIDs)
energy
harvesting
(EH),
multi-edge
servers,
one
server
(CS)
a
environment.
The
optimization
problem
formulated
minimize
average
long-term
service
cost
defined
as
weighted
sum
MID
consumption
delay,
constraints
available
resource,
causality,
allowable
maximum
decoding
error
probability.
address
involving
both
discrete
continuous
variables,
propose
multi-device
hybrid
decision-based
deep
reinforcement
learning
(DRL)
solution,
named
DDPG-D3QN
algorithm,
where
deterministic
policy
gradient
(DDPG)
dueling
double
Q
networks
(D3QN)
are
invoked
tackle
action
domains,
respectively.
Specifically,
improve
actor-critic
structure
DDPG
by
combining
D3QN.
It
utilizes
actor
part
search
for
optimal
rate
power
control
local
execution.
Meanwhile,
combines
critic
D3QN
select
offloading.
Simulation
results
demonstrate
proposed
algorithm
has
better
stability
faster
convergence,
while
achieving
higher
rewards
than
existing
DRL-based
methods.
Furthermore,
approach
outperforms
non-collaborative
schemes.
