Implementasi Full State Feedback LQR dengan JST pada Kendali Ketinggian Quadrotor

Abstract

One type of unmanned aircraft that is often used today is quadrotor. This type of aircraft has the ability to take off vertically. This study implemented an altitude control system on the z-axis quadrotor. The control used is the full state method of Linear Quadratic Regulator (LQR) with artificial neural networks. The LQR full state feedback method used in this system is 12-states with each feedback constant K tuned to the neural network method. This study implements the artificial neural network method to change the feedback constant on the z-axis. Artificial neural network architecture used 12 input layers, 48 hidden layers, and 1 output layer. This study compares the value of the results of the simulation with the response value of the system implementation results applied to the quadrotor. Testing with full state LQR feedback using artificial neural networks improves the system response to ±0.77 seconds and improves steady state error values up to ±12 cm. Based on the results of these studies, this system can be implemented to control other systems.