Electrical and Computer Engineering ETDs

Publication Date

Summer 7-15-2022

Abstract

In this paper, we propose an approach for shared control of a planar quadrotor that allows for non-Gaussian disturbance in the model and non-Gaussian variation in the pilot's control actions. We do this by constructing empirical characteristic functions for the state, inputs, and disturbance using demonstrations by a human expert. These are then used to make predictions of future states and of the system disturbance, using the first and second moments of the empirical characteristic function to estimate the mean and variance of these processes. With this method, we can extend assumptions from additive white Gaussian noise to any real-valued disturbance for a system with a quadratic cost. The proposed method is shown to properly find a control scheme and maintain system performance with a generalized disturbance model.

Keywords

stochastic control, shared control, optimal control, Non-Gaussian Processes

Document Type

Thesis

Degree Name

Electrical Engineering

Level of Degree

Masters

Department Name

Electrical and Computer Engineering

First Committee Member (Chair)

Meeko M. K. Oishi

Second Committee Member

Rafael Fierro

Third Committee Member

Claus Danielson

Download

Share

COinS