DEEP LEARNING BASED FLY-OVER WAYPOINTS CONTROL SYSTEM FOR BUSINESS JET AIRCRAFT

• Main Author: Kelvianto, Adi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/69001
• Institution: Institut Teknologi Bandung

In order to reach the intended destination, Area Navigation (RNAV) requires pilot to use fly-by and fly-over waypoints method. System that developed in this thesis is focusing on implementing fly-over waypoints method. PID controller strategies are commonly found in building this control system. However, by seeing the advantages offered by Deep Learning (DL) which can overcome problems that occurs in PID strategy. Hence, this thesis aims to develop DL-based fly over waypoints control system for Cirrus Vision SF50 aircraft and study the relationship between learning data characteristic with the result of control performance. Reconstruction of flight mission data through flight simulator that integrated with PID-based fly-over waypoints method is firstly done before creating DL model. The result obtained through this research are several DL-based fly-over control systems are able to provide a better balance of minimum distance to waypoints and cross track distances than the PID controller method. It was also found that the control characteristic of the DL model is closely related to the characteristic of the data used to train the model.