東華大學圖書館 |

Robust Machine Learning for Vision-based Navigation About Non-Cooperative Resident Space Objects /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Robust Machine Learning for Vision-based Navigation About Non-Cooperative Resident Space Objects // Tae Ha Park.
Author:	Park, Tae Ha,
Description:	1 electronic resource (210 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-03, Section: B.
Contained By:	Dissertations Abstracts International86-03B.
Subject:	Sensitivity analysis. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31460264
ISBN:	9798384340362

Robust Machine Learning for Vision-based Navigation About Non-Cooperative Resident Space Objects /
Park, Tae Ha,

Robust Machine Learning for Vision-based Navigation About Non-Cooperative Resident Space Objects /Tae Ha Park. - 1 electronic resource (210 pages)

Source: Dissertations Abstracts International, Volume: 86-03, Section: B.

As the space industry grows and more platforms are launched into outer space, there is a rising concern for its sustainable development and an immediate need for strategies to prevent potential crises. These strategies include various upcoming missions such as on-orbit servicing (e.g., inspection, refueling) and active debris removal which commonly require a safe and autonomous Rendezvous, Proximity Operation and Docking (RPOD) capability of a servicer spacecraft with respect to an arbitrary Resident Space Object (RSO). The core of RPOD is the real-time estimation and tracking of the position and orientation (i.e., pose) of the target RSO, and achieving this using a passive optical sensor such as a camera is particularly attractive as cameras are ubiquitous, low Size-Weight-Power-Cost (SWaP-C) sensors present on all satellite platforms. Machine Learning (ML) is a modern approach to such a complex task as predicting the 6D pose of an object from a single or sequence of images. Unfortunately, space is a challenging environment to operate data-driven ML approaches. First, the inaccessibility to space prevents collecting a large-scale image dataset for not only training but also validation of spaceborne ML methods. Second, the computational resource is limited onboard satellite avionics due to the harsh space environment. These two challenges demand a computationally efficient ML-based pose estimation and navigation algorithm and, more importantly, a mechanism by which the algorithm's performance and robustness on unavailable spaceborne flight images can be validated on-groundprior to deployment. In response, this dissertation makes numerous contributions to the state of the art for vision-based pose estimation of a non-cooperative target using ML. The proposed algorithms and methodologies allow for training and on-ground validation of Neural Networks (NN) and ML-in-the-loop navigation filters that are also computationally efficient when tested on representative satellite avionics. The first contribution is various open-source benchmark datasets designed for training and evaluating various NN models for spaceborne applications. All datasets commonly employ synthetic images rendered with computer graphics for NN training. Select datasets additionally consist of Hardware-In-the-Loop (HIL) images of a mockup satellite model captured with a real optical sensor in the Testbed for Rendezvous and Optical Navigation (TRON) facility at Stanford's Space Rendezvous Laboratory. Analyses on nearly 10,000 images from TRON reveal that HIL images are viable on-ground surrogates for otherwise unavailable flight images that can be used for pre-flight performance verification of ML-in-the-loop guidance, navigation and control algorithms.The second contribution is Spacecraft Pose Network v3 (SPNv3), a computationally efficient NN model for monocular pose estimation of a known target RSO, and a robust Adaptive Unscented Kalman Filter (AUKF) that utilizes SPNv3 as its measurement module. First, SPNv3 is a Vision Transformer (ViT)-based architecture which is trained exclusively on synthetic images of the target to predict heatmaps around known keypoints of its surfaces. Extensive trade-off analyses are performed to identify key aspects-lack of convolution operations, transfer learning and data augmentation-that contribute to achieving state-of-the-art OOD robustness while minimizing incurred computational overhead. A mid-size SPNv3 model can perform pose estimation at over 30Hz on the GPU of an NVIDIA Jetson Nano which is well above the nominal update frequency of modern satellite GN&C systems. Then, this SPNv3 is incorporated into the AUKF which tracks the relative orbital and attitude motion of the target. In order to account for the outlier measurements from SPNv3 due to the domain gap, the AUKF employs adaptive state noise compensation to solve for an optimal, positive semi-definite process noise covariance matrix based on the underlying continuous-time dynamics. Specifically, a novel analytical model for the process noise of relative attitude dynamics is newly derived and implemented. Experiments on HIL images of representative RPO scenarios suggest that the proposed AUKF is robust to measurement outliers from even the smallest SPNv3 variant with inferior performance.

English

ISBN: 9798384340362Subjects--Topical Terms:

3560752
Sensitivity analysis.

Robust Machine Learning for Vision-based Navigation About Non-Cooperative Resident Space Objects /
LDR:05717nmm a22003973i 4500 001 2438234
005 20251210113950.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798384340362
035 $a (MiAaPQD)AAI31460264
035 $a (MiAaPQD)Stanfordfk941fj3586
035 $a AAI31460264
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Park, Tae Ha, $e author. $3 3829477
245 1 0 $a Robust Machine Learning for Vision-based Navigation About Non-Cooperative Resident Space Objects / $c Tae Ha Park.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (210 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-03, Section: B.
500 $a Advisors: D'Amico, Simone; Kochenderfer, Mykel; Schwager, Mac.
502 $b Ph.D. $c Stanford University $d 2024.
520 $a As the space industry grows and more platforms are launched into outer space, there is a rising concern for its sustainable development and an immediate need for strategies to prevent potential crises. These strategies include various upcoming missions such as on-orbit servicing (e.g., inspection, refueling) and active debris removal which commonly require a safe and autonomous Rendezvous, Proximity Operation and Docking (RPOD) capability of a servicer spacecraft with respect to an arbitrary Resident Space Object (RSO). The core of RPOD is the real-time estimation and tracking of the position and orientation (i.e., pose) of the target RSO, and achieving this using a passive optical sensor such as a camera is particularly attractive as cameras are ubiquitous, low Size-Weight-Power-Cost (SWaP-C) sensors present on all satellite platforms. Machine Learning (ML) is a modern approach to such a complex task as predicting the 6D pose of an object from a single or sequence of images. Unfortunately, space is a challenging environment to operate data-driven ML approaches. First, the inaccessibility to space prevents collecting a large-scale image dataset for not only training but also validation of spaceborne ML methods. Second, the computational resource is limited onboard satellite avionics due to the harsh space environment. These two challenges demand a computationally efficient ML-based pose estimation and navigation algorithm and, more importantly, a mechanism by which the algorithm's performance and robustness on unavailable spaceborne flight images can be validated on-groundprior to deployment. In response, this dissertation makes numerous contributions to the state of the art for vision-based pose estimation of a non-cooperative target using ML. The proposed algorithms and methodologies allow for training and on-ground validation of Neural Networks (NN) and ML-in-the-loop navigation filters that are also computationally efficient when tested on representative satellite avionics. The first contribution is various open-source benchmark datasets designed for training and evaluating various NN models for spaceborne applications. All datasets commonly employ synthetic images rendered with computer graphics for NN training. Select datasets additionally consist of Hardware-In-the-Loop (HIL) images of a mockup satellite model captured with a real optical sensor in the Testbed for Rendezvous and Optical Navigation (TRON) facility at Stanford's Space Rendezvous Laboratory. Analyses on nearly 10,000 images from TRON reveal that HIL images are viable on-ground surrogates for otherwise unavailable flight images that can be used for pre-flight performance verification of ML-in-the-loop guidance, navigation and control algorithms.The second contribution is Spacecraft Pose Network v3 (SPNv3), a computationally efficient NN model for monocular pose estimation of a known target RSO, and a robust Adaptive Unscented Kalman Filter (AUKF) that utilizes SPNv3 as its measurement module. First, SPNv3 is a Vision Transformer (ViT)-based architecture which is trained exclusively on synthetic images of the target to predict heatmaps around known keypoints of its surfaces. Extensive trade-off analyses are performed to identify key aspects-lack of convolution operations, transfer learning and data augmentation-that contribute to achieving state-of-the-art OOD robustness while minimizing incurred computational overhead. A mid-size SPNv3 model can perform pose estimation at over 30Hz on the GPU of an NVIDIA Jetson Nano which is well above the nominal update frequency of modern satellite GN&C systems. Then, this SPNv3 is incorporated into the AUKF which tracks the relative orbital and attitude motion of the target. In order to account for the outlier measurements from SPNv3 due to the domain gap, the AUKF employs adaptive state noise compensation to solve for an optimal, positive semi-definite process noise covariance matrix based on the underlying continuous-time dynamics. Specifically, a novel analytical model for the process noise of relative attitude dynamics is newly derived and implemented. Experiments on HIL images of representative RPO scenarios suggest that the proposed AUKF is robust to measurement outliers from even the smallest SPNv3 variant with inferior performance.
546 $a English
590 $a School code: 0212
650 4 $a Sensitivity analysis. $3 3560752
650 4 $a Satellites. $3 924316
650 4 $a Vision systems. $3 3685322
650 4 $a Orbits. $3 534192
650 4 $a Neural networks. $3 677449
650 4 $a Aerospace engineering. $3 1002622
690 $a 0538
690 $a 0800
710 2 $a Stanford University. $e degree granting institution. $3 3765820
720 1 $a D'Amico, Simone $e degree supervisor.
720 1 $a Kochenderfer, Mykel $e degree supervisor.
720 1 $a Schwager, Mac $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-03B.
790 $a 0212
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31460264