Electrical and Computer Engineering ETDs

Publication Date

Spring 4-13-2023

Abstract

Stochastic disturbances arise in a variety of engineering applications. For tractability, Gaussian disturbances are often assumed. However, this may not always be valid, such as when a disturbance exhibits heavy-tailed or skewed phenomena. As autonomous systems become more ubiquitous, non-Gaussian disturbances will become more common due to the compounding effects of sensing, actuation, and external forces. Despite this, little has been done to develop formal methods that are both computationally efficient and allow for analytical assurances with non-Gaussian disturbances. Addressing convex polytopic set acquisition and non-convex collision avoidance chance constraints with quantile and moment-based reformulations, this dissertation proposes novel stochastic optimal control techniques that are computationally efficient and allow for analytic guarantees with arbitrary disturbances. These reformulations are amenable to optimization techniques while eliminating costly, and frequently intractable, high-dimensional integrals. These conservative reformulations guarantee chance constraint satisfaction and are numerically tractable. I demonstrate these methods with applications to multi-satellite operations.

Keywords

Stochastic optimal control, chance constraints, Control theory, Optimization, Stochastic motion planning

Document Type

Dissertation

Language

English

Degree Name

Electrical Engineering

Level of Degree

Doctoral

Department Name

Electrical and Computer Engineering

First Committee Member (Chair)

Meeko Oishi

Second Committee Member

Rafael Fierro

Third Committee Member

Claus Danielson

Fourth Committee Member

Christopher Petersen

Download

Share

COinS