Masters Theses

Date of Award

8-2025

Degree Type

Thesis

Degree Name

Master of Science

Major

Electrical Engineering

Major Professor

Garrett Rose

Committee Members

Catherine Schuman, James Plank

Abstract

Embedded neuromorphic applications propose the unique blend of analog computational practices within the more classically digital embedded framework. The neural network itself has made headway into numerous fields including data analytics and self-driving vehicles. It offers a robust processing capability for multi-variable and multi-state implementations. Neural networks present a novel approach to solving problems of scaling size. Each input sensor receives its own weights, which each layer of the network may interact with, ultimately coalescing in a single output collectively decided by the network. In conjunction, embedded applications have provided numerous solutions for Internet of Things (IoT) applications. These include small-scale circuits regulated temperature to more advanced traffic-signal modules. Usually, these boil down to what is known as a ”control” application. Some input amends the current state of the system, and the central controller attempts to return to a homeostasis. This work attempts to utilize a neuromorphic framework in place of a more typical digital controller in order to localize image processing and control. The controller will operate a camera and a remotely-controlled robot in response to what it identifies on camera.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS