Using a UAV and Edge Computing to Identify and Throw Away Trash
Open Access
Author:
Eden, Grant
Graduate Program:
Computer Science and Engineering
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
March 29, 2020
Committee Members:
Vijaykrishnan Narayanan, Thesis Advisor/Co-Advisor John Morgan Sampson, Thesis Advisor/Co-Advisor Bhuvan Urgaonkar, Program Head/Chair
Keywords:
UAV Edge Computing Drone TPU Machine Learning
Abstract:
This thesis explores the application of edge computing on unmanned aircraft vehicle to provide autonomous trash identification and disposal. The main focus is finding a way to run real time object detection onboard the UAV without having to stream data to a remote server. Few works explore the use of accelerators on drones to increase machine learning inference computational speed. With this, this thesis explores the quantization of neural networks to provide lightweight and accurate object detection models. This thesis also explores the creation of an image set to train a model for trash detection, specifically plastic bags, plastic bottles, and cans. The thesis demonstrates the application for autonomous, quick, and easy trash disposal via drone and the potential for commercial use of drones for environmental cleanup.
