Date on Master's Thesis/Doctoral Dissertation

5-2016

Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Department

Computer Engineering and Computer Science

Degree Program

Computer Science and Engineering, PhD

Committee Chair

Lauf, Adrian

Committee Co-Chair (if applicable)

Kantardzic, Mehmed

Committee Member

Kantardzic, Mehmed

Committee Member

Yampolskiy, Roman V.

Committee Member

Li, Hongxiang

Committee Member

Welch, Karla

Author's Keywords

link quality estimation; wireless robot networks; online fuzzy learning of communication quality

Abstract

It is often essential for robots to maintain wireless connectivity with other systems so that commands, sensor data, and other situational information can be exchanged. Unfortunately, maintaining sufficient connection quality between these systems can be problematic. Robot mobility, combined with the attenuation and rapid dynamics associated with radio wave propagation, can cause frequent link quality (LQ) issues such as degraded throughput, temporary disconnects, or even link failure. In order to proactively mitigate such problems, robots must possess the capability, at the application layer, to gauge the quality of their wireless connections. However, many of the existing approaches lack adaptability or the framework necessary to rapidly build and sustain an accurate LQ prediction model. The primary contribution of this dissertation is the introduction of a novel way of blending machine learning with fuzzy logic so that an adaptable, yet intuitive LQ prediction model can be formed. Another significant contribution includes the evaluation of a unique active and incremental learning framework for quickly constructing and maintaining prediction models in robot networks with minimal sampling overhead.

Download

Share

COinS