Design and flight evaluation of deep model predictive control

McCann, Dennis

Permalink

https://hdl.handle.net/2142/124381 Copy

Description

Title

Design and flight evaluation of deep model predictive control

Author(s)

McCann, Dennis

Issue Date

2024-04-30

Director of Research (if dissertation) or Advisor (if thesis)

Chowdhary, Girish

Department of Study

Aerospace Engineering

Discipline

Aerospace Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

Deep Model Predictive Control, Deep Learning, Adaptive Control, Online Learning, Model Predictive Control, Quadrotors

Abstract

As autonomous systems are increasingly employed in our society, the necessity for robust and adaptive control mechanisms to ensure their safety and effectiveness has become paramount. This thesis focuses on the implementation and evaluation of Deep Model Predictive Control, an adaptive control algorithm designed to manage nonlinear systems experiencing matched and bounded state-dependent uncertainties or disturbances. Deep Model Predictive Control stands out for its ability to make real-time adjustments to disturbances, utilizing a deep learning based adaptive architecture, while satisfying system constraints and stability guarantees. Through a comprehensive series of experiments escalating in complexity—from numerical simulations to flight tests on actual hardware—this research aims to validate the efficacy and applicability of Deep Model Predictive Control in enhancing the autonomy, reliability, and safety of these systems. By systematically increasing the fidelity of our experimental evaluations, not only the theoretical viability of Deep Model Predictive Control is assessed, but also the practical viability, by implementing the algorithm to control the physical Crazyflie 2.0 quadrotor. This work endeavors to bridge theoretical adaptive control strategy of Deep Model Predictive Control with real-world applications, marking a step forward in the deployment of adaptive controllers that harness deep learning for the estimation and mitigation of uncertainty.

Graduation Semester

2024-05

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/124381

Copyright and License Information

Copyright 2024 Dennis McCann

Owning Collections