Abstract: In recent years,
unmanned aerial
vehicleS (UAVs) or
also known as
drones have been
widely used for
military missions and
commercial
applications. The
number of practical
applications of UAV
has been increased
dramatically because
of utilizing a wide
variety of research
fields such as
Artificial Intelligence
(AI), image
processing, and
autonomous control.
In this work, we
present the design of
the guidance system
of autonomous UAVs
for target tracking.
The path planning is
crucial for
controlling UAVs to
track target. We
design the path
planning and target
tracking problem by
partitioning into two
major subproblems,
multiple targets
assignment and
path planning. The
target assignment
problem is solved
using Hybrid Genetic
Algorithm (HGA)
and the path
planning problem is
solved using the
belief Monte Carlo
Tree Search (MCTS)
with Double Q-
learning
methodology. The
path planning and
target tracking
problem are
formulated as
Partially Observable
Markov Decision
Process (POMDP)
problem. The
POMDP with long-
term decision
support systems can
improve the
inaccuracy of limited
resources
management and
long term prediction
with noise
interference. Using
the belief MCTS with
Double Q-learning
approach
incorporated with
nominal belief-state
optimization (NBO)
method for solving
POMDP problem is a
practical approach to
the autonomous
UAVs path planning
for target tracking
problem. The
multiple targets
assignment using
Hybrid Genetic
Algorithm is able to
increase the
efficiency of UAVs
flight path planning.
The simple genetic
operators are
selection, crossover,
and mutation. For
crossover, we
develop a new
operator, greedy
crossover which is
called HGA operator
by using domain-
specic knowledge of
the target
assignment problem.
We also propose the
process of
generating an initial
small population for
selection operator.
The belief MCTS
path planning
algorithm and target
tracking are
successful in solving
the moving target
tracking problem
with the UAVs path
planning
management.

Adviser: Prof. Edwin K.P. Chong
Co-Adviser: N/A
Non-ECE Member: Prof. Olivier Pinaud
Member 3: Prof. Mahmood R Azimi-Sadjadi
Addional Members: Prof. Ali Pezeshki

Publications:
The UAVs Path Planning using Monte Carlo Tree Search method

Program of Study:
ECE 513
ECE 520
ECE 656
ECE 681A1
ECE 516
ECE 514
N/A
N/A