Přístupnostní navigace
E-application
Search Search Close
Publication detail
NOVÁK, J. CHUDÝ, P.
Original Title
Dynamic Soaring in Uncertain Wind Conditions: Polynomial Chaos Expansion Approach
Type
conference paper
Language
English
Original Abstract
Dynamic soaring refers to a flight technique used primarily by large seabirds to extract energy from the wind shear layers formed above ocean surface. A small Unmanned Aerial Vehicle (UAV) capable of efficient dynamic soaring maneuvers can enable long endurance missions in context of patrol or increased flight range. To realize autonomous energy-saving patterns by a UAV, a real-time trajectory generation for a dynamic soaring maneuver accounting for varying external conditions has to be performed. The design of the flight trajectory is formulated as an Optimal Control Problem (OCP) and solved within direct collocation based optimization. A surrogate model of the optimal traveling cycle capturing wind profile uncertainties is constructed using Polynomial Chaos Expansion (PCE). The unknown wind profile parameters are estimated from observed trajectory by means of a Genetic Algorithm (GA). The PCE surrogate model is subsequently utilized to update the optimal trajectory using the estimated wind profile parameters.
Keywords
Polynomial Chaos Expansion, Surrogate Modeling, Dynamic Soaring, Optimal Control
Authors
NOVÁK, J.; CHUDÝ, P.
Released
16. 2. 2024
Publisher
Springer Nature Switzerland AG
Location
Grasmere
ISBN
978-3-031-53968-8
Book
Machine Learning, Optimization, and Data Science
Edition
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
0302-9743
Periodical
Lecture Notes in Computer Science
Number
14505
State
Federal Republic of Germany
Pages from
104
Pages to
115
Pages count
11
BibTex
@inproceedings{BUT185184, author="Jiří {Novák} and Peter {Chudý}", title="Dynamic Soaring in Uncertain Wind Conditions: Polynomial Chaos Expansion Approach", booktitle="Machine Learning, Optimization, and Data Science", year="2024", series="Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)", journal="Lecture Notes in Computer Science", number="14505", pages="104--115", publisher="Springer Nature Switzerland AG", address="Grasmere", doi="10.1007/978-3-031-53969-5\{_}9", isbn="978-3-031-53968-8", issn="0302-9743" }