Predictive modelling of quadrotor crash trajectory subjected to single and dual motor failures

Unmanned aerial vehicles (UAVs), specifically quadcopters, have seen a surge in utilization, playing critical roles in enhancing efficiency and effectiveness in the recreational and commercial sectors. However, despite its potential, quadcopters tend to have a higher failure rate, the most common be...

Full description

Saved in:
Bibliographic Details
Main Author: Yap, Nicholas Kit
Other Authors: Mir Feroskhan
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2024
Subjects:
Online Access:https://hdl.handle.net/10356/177495
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Unmanned aerial vehicles (UAVs), specifically quadcopters, have seen a surge in utilization, playing critical roles in enhancing efficiency and effectiveness in the recreational and commercial sectors. However, despite its potential, quadcopters tend to have a higher failure rate, the most common being propulsion failure. Although the public has recognised this flaw, the propulsion failure of a quadcopter has been understudied, with most available works of literature focused solely on the Complete Power Failure of a Quadcopter. To bridge this gap, RFlySim, which integrates MATLAB/Simulink and Pixhawk/PX4 software, has been utilised to conduct a comprehensive analysis of factors, including fluctuating thrust losses across various masses, failure time, and initial velocities to understand the crash trajectories of a quadcopter. In this study, both statistical and visual analyses were conducted. Simulation outcomes indicate that a hovering quadcopter with a Single Motor Failure has a crash radius of 10-meters whereas incorporating an initial velocity to a quadcopter result in a crash radius that is four times more. On the other hand, when Dual Rotor Failure is induced at a hovering state, crash points are predominantly concentrated at the origin, whereas at instances when an initial velocity was introduced, a higher concentration of crash occurrences occurred within the 15-to-40-meter zone. Beyond the simulation, prediction modelling was carried out with the LSTM model displaying superiority over other models in predicting the quadcopters’ crash trajectory with an MAE and RMSE value of 0.0146 and 0.0181.