Graduate Thesis Or Dissertation
 

Learning-based control and coordination of autonomous UAVs

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/2227ms09q

Descriptions

Attribute NameValues
Creator
Abstract
  • Uninhabited aerial vehicles, also called UAVs are currently controller by a combination of a human pilot at a remote location, and autopilot systems similar to those found on commercial aircraft. As UAVs transition from remote piloting to fully autonomous operation, control laws must be developed for the tasks to be performed. Flight control and navigation are low-level tasks that must be performed by the UAV in order to complete more useful missions. In the domain of persistent aerial surveillance, in which UAVs are responsible for locating and continually observing points of interest (POIs) in the environment, such a mission can be accomplished much more efficiently by groups of cooperating UAVs. To develop the controller for a UAV, a discrete-time, physics-based simulator was developed in which an initially random neural network controller could be evolved over successive generations to produce the desired output. Because of the inherent complexity of navigating and maintaining stable flight, a novel state space utilizing an approximation of the flight path length between the aircraft and its navigational waypoint is developed and implemented. In choosing the controller output as the net thrust of the aircraft from all control surfaces and impellers, a controller suitable for a wide range of UAV types is reached. To develop a controller for each aircraft to cooperate in the persistent aerial surveillance domain, a behavior-based simulator was developed. Using this simulator, constraints on the flight dynamics are approximated to speed computation. Each UAV agent trains a neural network controller through successive episodes using sensory data about other aircraft and POIs. Testing of each controller was done by simulating in increasingly dynamic environments. The flight controller is shown to be able to successfully maintain heading and altitude and to make turns to ultimately reach a waypoint. The surveillance coordination controller is shown to coordinate UAVs well for both static and mobile POIs, and to scale well from systems of 3 agents to systems of 30 agents. Scaling of the controller to more agents is particularly effective when using a difference reward calculation in training the controllers.
License
Resource Type
Date Available
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Non-Academic Affiliation
Subject
Rights Statement
Publisher
Peer Reviewed
Language
Replaces

Relationships

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details bell_thesis_dec9.pdf 2017-08-07 Public Download