Date of Award

5-2024

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Mechanical and Civil Engineering

First Advisor

Hector M. Gutierrez

Second Advisor

Eric D. Swenson

Third Advisor

Seong Hyeon Hong

Fourth Advisor

Ashok Pandit

Abstract

Comprehensive research on Unmanned Aerial Vehicles (UAV) system identification for motion control parameters is presented in this thesis, with a focus on the necessity of precise control and improved performance. Using Pseudorandom Binary Sequence (PRBS) and Normally Distributed Random Numbers, it presents a unique technique for excitation of UAV dynamic systems. It also shows how effective random signals are in time-domain identification for precise control in a range of flying circumstances. The piece of research includes a thorough analysis and implementation of various approaches and its further improvements, highlighting the benefits and drawbacks of each. These approaches include the free flight manual three degrees of freedom excitation of all vehicle axes, the PRBS input excitation and further development.

In order to enhance system identification, the thesis explores extensively into the implementation of ROS 2-PX4, using current UAV control technology instead of ROS-MAVROS-PX4. With the goal of obtaining a rich dataset for model estimation, it presents a unique approach for system identification from motion commands in Gazebo. It does this by utilizing PRBS and random noise inputs to reproduce realistic UAV dynamics without limiting the degrees of freedom.

The thesis describes a methodologically rigorous procedure that includes setting up ROS 2 nodes in Gazebo UAV simulator, pre-processing data, selecting and assessing models, and validating the results, comparing through models acquired from free flight testing on an actual quadrotor. It provides an extensive overview of the transfer functions and controller parameters acquired for the roll, pitch, and yaw axes, demonstrating the accuracy of the suggested models through simulated responses and frequency response comparisons.

The outcome of the research confirms the superiority of the suggested system identification techniques over traditional approaches, demonstrating how random noise inputs may be used to capture a broad range of UAV dynamics. The validated models provide considerable improvements to the design of UAV control systems, facilitating the development of more responsive and advanced motion control algorithms such as Model predictive control (MPC). By highlighting the crucial role that accurate system identification plays in improving UAV performance and adaptability, this analysis not only advances the area of UAV system identification but also paves the way for future research into complex control systems.

Download

Available for download on Sunday, May 04, 2025

Share

COinS