A rule-based dialog management system for integration of unmanned aerial systems into the national airspace system
Abstract/Contents
- Abstract
- Unmanned Aerial Systems (UAS) are one of the fastest growing segments of the aviation industry with the U.S. UAS market projected to nearly double by 2020. UAS have been used extensively overseas in military roles, but there are many civilian missions that could be performed by UAS operating domestically such as law enforcement, homeland security, wildfire suppression, air-freight, and aerial-survey. However, despite these potential markets, UAS are currently banned from flying within the U.S. National Airspace System (NAS) by federal regulation. This ban is due primarily to reasons related to operational safety of flight, among them the inability of UAS to respond to and engage in the spoken Air Traffic Control (ATC) instructions that govern the use of the NAS. In this thesis an Autonomous Communication System (AutoComm System) and Dialog Management System (DMS) are developed that enables UAS to autonomously communicate with ATC using spoken natural language speech. The AutoComm System uses probabilistic techniques to automatically learn the rules of ATC communications and models the dialog as a Bayesian network. In order to implement the AutoComm System several novel contributions were developed. These include: a novel scheme for part of speech tagging for ATC dialog; a central sliding window algorithm for determining the most probable part-of-speech tagging in the presence of noise and transcription errors in the sentence being labeled; and a robust hybrid method for enabling a DMS to automatically learn rule-based dialog trees while accounting for non-deterministic dialog and variance in accepted phraseology. The approach discussed in this thesis is formulated to run on the potentially limited computing power found on smaller UAS while still enabling them to speak with the very high accuracy necessary for operation in the NAS. The results of flying the AutoComm System through simulated and real-world ATC scenarios show that the AutoComm System can correctly respond to ATC communications with an accuracy of 88.5%. This lags the observed 98.9% accuracy rate of human pilots. However, this performance gap is caused almost entirely by scenarios where ATC failed to properly self-identify their transmissions as required by the FAA. It is possible for the AutoComm System developed in this thesis to approach a human-equivalent level of performance with more rigorous ATC adherence to FAA phraseology and a more extensive training corpus. Further, in its current form, the AutoComm System can outperform human pilots in a few key areas such as the ability to not interrupt inappropriately and proper phrasing of messages. These results show that it is possible to develop and implement a high-accuracy system that enables UAS to autonomously communicate with ATC in order to better integrate UAS into the NAS for a wide variety of missions.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2012 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Gunawardana, Sandun Uthpala |
|---|---|
| Associated with | Stanford University, Department of Aeronautics and Astronautics |
| Primary advisor | Alonso, Juan José, 1968- |
| Thesis advisor | Alonso, Juan José, 1968- |
| Thesis advisor | Erzberger, Heinz |
| Thesis advisor | Pavone, Marco, 1980- |
| Advisor | Erzberger, Heinz |
| Advisor | Pavone, Marco, 1980- |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | San Gunawardana. |
|---|---|
| Note | Submitted to the Department of Aeronautics and Astronautics. |
| Thesis | Ph.D. Stanford University 2012 |
| Location | electronic resource |
Access conditions
- Copyright
- © 2012 by Sandun Uthpala Gunawardana
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...