Trajectory Tracking Control of ISR Quadrotor UAV using GA Optimized Fuzzy-PID Based Neural Network Controller
No Thumbnail Available
Date
2023-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
The ISR Quad-Rotor is an unmanned aerial vehicle (UAV) utilized for intelligence gathering,
surveillance, and reconnaissance missions. This thesis focuses on modeling and controlling
the quad-rotor to identify and track a Person of interest using an intelligent control technique.
The initial step involves deriving a nonlinear mathematical model for the 6DOF (six degrees
of freedom) Quadrotor UAV using the Newton-Euler formalism. To ensure trajectory tracking
of the quad-rotor, a Fuzzy Proportional-Integral-Derivative (Fuzzy-PID) controller, tuned
with a Genetic Algorithm (GA), was implemented to generate data for a neural network
(NNFPID-GA). The fuzzy logic approach facilitates parameter adjustment based on predefined
fuzzy rules, while the GA algorithm determines the scaling factors of the Fuzzy-PID controller.
The proposed control system was designed based on input-output data from the GA Optimized
Fuzzy-PID controller. A network was trained using the Levenberg-Marquardt backpropagation
algorithm with the assistance of the MATLAB®NN Toolbox. MATLAB®simulations were
conducted to validate the effectiveness of the proposed control algorithm. Additionally, a
flight test were performed using the UAV Toolbox to assess the stable flight performance
of a developed GA Optimized Fuzzy-PID based neural network(NNFPID-GA) controller for
the Quadrotor UAV. To evaluate the performance of both controllers, a comparison study
based on performance metrics was conducted. Even though both controllers offer faster
response in terms of settling time and improved performance, with less overshoot and better
robustness in handling parameter variation and disturbance rejection capability, the GA
Optimized Fuzzy-PID based neural network controller(NNFPID-GA) outperforms the GA
Optimized Fuzzy-PID controller(FPID-GA). Finally, in order to generate trajectories for the
controller, a face recognition system using Python were implemented. The training results
demonstrate an accuracy of 98.65%, indicating that the system can effectively distinguish
between a wanted person (Person of interest) and other individuals.
Description
Keywords
Quad-Rotor, Neural Network, FPID-GA,NNFPID-GA, Genetic Algorithm, Degrees of Freedom, ISR, UAV.