ISPRS International Journal of Geo-Information,
Journal Year:
2024,
Volume and Issue:
13(8), P. 264 - 264
Published: July 26, 2024
This
study
introduces
an
innovative
scheme
for
classifying
uncrewed
aerial
vehicle
(UAV)-derived
trajectory
behaviors
by
employing
machine
learning
(ML)
techniques
to
transform
original
trajectories
into
various
sequences:
space–time,
speed–time,
and
azimuth–time.
These
transformed
sequences
were
subjected
normalization
uniform
data
analysis,
facilitating
the
classification
of
six
distinct
categories
through
application
three
ML
classifiers:
random
forest,
time
series
forest
(TSF),
canonical
characteristics.
Testing
was
performed
across
different
intersections
reveal
accuracy
exceeding
90%,
underlining
superior
performance
integrating
azimuth–time
speed–time
over
conventional
space–time
analyzing
behaviors.
research
highlights
TSF
classifier’s
robustness
when
incorporating
speed
data,
demonstrating
its
efficiency
in
feature
extraction
reliability
intricate
pattern
handling.
study’s
results
indicate
that
direction
information
significantly
enhances
predictive
model
robustness.
comprehensive
approach,
which
leverages
UAV-derived
advanced
techniques,
represents
a
significant
step
forward
understanding
behaviors,
aligning
with
goals
enhancing
traffic
control
management
strategies
better
urban
mobility.
Frontiers in Neurorobotics,
Journal Year:
2025,
Volume and Issue:
19
Published: Jan. 23, 2025
Traffic
forecasting
is
crucial
for
a
variety
of
applications,
including
route
optimization,
signal
management,
and
travel
time
estimation.
However,
many
existing
prediction
models
struggle
to
accurately
capture
the
spatiotemporal
patterns
in
traffic
data
due
its
inherent
nonlinearity,
high
dimensionality,
complex
dependencies.
To
address
these
challenges,
short-term
model,
Trafficformer,
proposed
based
on
Transformer
framework.
The
model
first
uses
multilayer
perceptron
extract
features
from
historical
data,
then
enhances
spatial
interactions
through
Transformer-based
encoding.
By
incorporating
road
network
topology,
mask
filters
out
noise
irrelevant
interactions,
improving
accuracy.
Finally,
speed
predicted
using
another
perceptron.
In
experiments,
Trafficformer
evaluated
Seattle
Loop
Detector
dataset.
It
compared
with
six
baseline
methods,
Mean
Absolute
Error,
Percentage
Root
Square
Error
used
as
metrics.
results
show
that
not
only
has
higher
accuracy,
but
also
can
effectively
identify
key
sections,
great
potential
intelligent
control
optimization
refined
resource
allocation.
Computer-Aided Civil and Infrastructure Engineering,
Journal Year:
2025,
Volume and Issue:
unknown
Published: April 28, 2025
Abstract
Fast
and
accurate
identification
of
traffic
anomalies
on
highways
is
utmost
importance.
This
study
presents
an
integrated
framework
for
multiple
anomaly
detection
using
vehicle
trajectories.
The
addresses
both
macroscopic
congestion
patterns
microscopic
driving
behaviors,
offering
a
comprehensive
solution
that
simultaneously
detects
within
unified
framework.
developed
comprises
three
main
components:
data
acquisition
preprocessing,
trajectory
recognition,
detection.
former
two
components
are
responsible
acquiring
real‐time
trajectories
highways.
With
such
information
the
continuously
monitored
short‐term
state,
latter
component
seeks
to
detect
all
via
tailored
sub‐algorithm
each
them.
For
detection,
algorithm
detecting
stop‐and‐go
waves
by
constructing
localized
shockwaves
proposed
capture
propagation
even
in
limited
field‐of‐view
scenarios.
dynamic
background
state
updating
mechanism
introduced,
allowing
adaptively
integrate
historical
environmental
factors.
Additionally,
double‐layer
stacking
based
unsupervised
methods
designed
diverse
feature
types
addressing
perspective
distortions.
tested
experiments
simulation
real‐world
results
confirm
its
effectiveness
simultaneous
Journal of Advanced Transportation,
Journal Year:
2025,
Volume and Issue:
2025(1)
Published: Jan. 1, 2025
The
traffic
flow
at
freeway
on‐ramps
is
influenced
not
only
by
the
lane
changes
made
merging
vehicles
but
also
longitudinal
behavior
of
and
in
main
lane.
Existing
car‐following
models
are
suitable
to
represent
during
because
they
based
on
idea
that
intend
reach
a
steady
state,
is,
constant
time
headway
zero
speed
difference,
as
soon
possible.
At
on‐ramps,
however,
have
this
state
until
end
on‐ramp.
We
therefore
derive
novel
model
desired
headways
able
continuous
adaptation
toward
state.
From
model,
we
change
for
with
seven
parameters.
includes
leader
selection
algorithm,
which
enables
pass
or
be
passed
components
predict
start
surrogate
safety
measures
describe
lateral
change.
Due
resemblance
models,
methodology
calibrate
microscopic
scale
can
adopted
from
models.
To
validate
conduct
simulations
compare
simulated
trajectory
data
two
German
on‐ramps.
results
show
accurately
represents
driving
their
followers,
although
it
slightly
overestimates
number
passing
vehicle
under
congested
conditions.
yield
accurate
distributions,
except
cases
very
risky
driver
behavior,
realistically
capture
macroscopic
speed‐density
relationship
