BROWSE

ITEM VIEW & DOWNLOAD

Sensor Fusion and Localization on Cooperative UAVs Formation

Cited 0 times inthomson ciCited 0 times inthomson ci
Title
Sensor Fusion and Localization on Cooperative UAVs Formation
Author
Kim, Myunggun
Advisor
Son, Hungsun
Issue Date
2022-02
Publisher
Ulsan National Institute of Science and Technology
Abstract
The automated formation system of unmanned aerial vehicles (UAVs) is expected to be utilized for many applications, such as investigation on harsh and broad area, delivery, target searching, or airborne operations like aerial refueling and repairing. However, it is rare that the applications are operated with the UAV formation system, because of lack of reliability and safety. This thesis is focused on the UAV formation system to be operated for sophisticate applications. Therefore, the mission target estimation, advanced UAV formation system estimation, and safety enhanced formation control are developed. The target estimation is as much as important to well operating the UAVs for successful mission achievement. Especially, the UAV system is sometimes required operating mission with low-cost sensors and poor information, so dynamics-based vehicle smoothing estimation (VSE) is developed. Data association technique maximize the usage of information and the dynamical characteristic of the target improves the performance of the algorithm. The dynamics-based fIPDA/bIPDA fixed lag smoother is applied to prevent track missing due to the radical movement of the target. Furthermore, the VSE evaluates estimated target candidate to distinguish the mission target among the false objects. The formation estimator provides the state and condition of the formation. The inevitable respective errors of individual UAVs, which might not be critical for each UAV, make vital formation structural error. Therefore, the formation restoring algorithm not only estimate the formation but also provides the solution to minimize the formation error. The state constrained extended Kalman filter-based sensor fusion is used to find best estimation minimizing bias error on position and restoring the shape of formation. The base structure of formation control is designed with the artificial potential field and leader-follower formation structure. Moreover, the formation control focuses on the safety, especially, for the interaction of UAVs. The UAVs on the formation affects each other and those interaction could degrade the entire performance, and they could crash each other in the extreme. Therefore, the adaptive repulsive potential with stochastic analysis for neighbors is proposed, so that the UAV could prevent the crash to the uncertain neighbor UAVs when they fly close to each other. Furthermore, the cooperative A3C disturbance observer is designed for more direct interaction between UAVs. The proposed reinforcement learning algorithm identifies the relationship between the neighbor maneuver and disturbance. The thesis discusses each topic with the major application, the anti-drone system, which estimate the hostile UAV and catch it down with automated UAV formation system.
Description
Department of Mechanical Engineering
URI
Go to Link
Appears in Collections:
MEN_Theses_Ph.D.
Files in This Item:
605511.pdf Download

find_unist can give you direct access to the published full text of this article. (UNISTARs only)

Show full item record

qrcode

  • mendeley

    citeulike

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

MENU