Details
| Originalsprache | Englisch |
|---|---|
| Titel des Sammelwerks | Proceedings of 2020 23rd International Conference on Information Fusion, FUSION 2020 |
| Herausgeber (Verlag) | Institute of Electrical and Electronics Engineers Inc. |
| ISBN (elektronisch) | 9780578647098 |
| Publikationsstatus | Veröffentlicht - Juli 2020 |
| Extern publiziert | Ja |
| Veranstaltung | 23rd International Conference on Information Fusion, FUSION 2020 - Virtual, Pretoria, Südafrika Dauer: 6 Juli 2020 → 9 Juli 2020 |
Publikationsreihe
| Name | Proceedings of 2020 23rd International Conference on Information Fusion, FUSION 2020 |
|---|
Abstract
Inertial measurement units are commonly used to estimate the attitude of moving objects. Numerous nonlinear filter approaches have been proposed for solving the inherent sensor fusion problem. However, when a large range of different dynamic and static rotational and translational motions is considered, the attainable accuracy is limited by the need for situation-dependent adjustment of accelerometer and gyroscope fusion weights. We investigate to what extent these limitations can be overcome by means of artificial neural networks and how much domain-specific optimization of the neural network model is required to outperform the conventional filter solution. A diverse set of motion recordings with a marker-based optical ground truth is used for performance evaluation and comparison. The proposed neural networks are found to outperform the conventional filter across all motions only if domain-specific optimizations are introduced. We conclude that they are a promising tool for inertial-sensor-based real-time attitude estimation, but both expert knowledge and rich datasets are required to achieve top performance.
ASJC Scopus Sachgebiete
- Informatik (insg.)
- Maschinelles Sehen und Mustererkennung
- Informatik (insg.)
- Information systems
- Entscheidungswissenschaften (insg.)
- Informationssysteme und -management
- Physik und Astronomie (insg.)
- Instrumentierung
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
Proceedings of 2020 23rd International Conference on Information Fusion, FUSION 2020. Institute of Electrical and Electronics Engineers Inc., 2020. 9190634 (Proceedings of 2020 23rd International Conference on Information Fusion, FUSION 2020).
Publikation: Beitrag in Buch/Bericht/Sammelwerk/Konferenzband › Aufsatz in Konferenzband › Forschung › Peer-Review
}
TY - GEN
T1 - Neural networks versus conventional filters for inertial-sensor-based attitude estimation
AU - Weber, Daniel
AU - Guhmann, Clemens
AU - Seel, Thomas
N1 - Publisher Copyright: © 2020 International Society of Information Fusion (ISIF).
PY - 2020/7
Y1 - 2020/7
N2 - Inertial measurement units are commonly used to estimate the attitude of moving objects. Numerous nonlinear filter approaches have been proposed for solving the inherent sensor fusion problem. However, when a large range of different dynamic and static rotational and translational motions is considered, the attainable accuracy is limited by the need for situation-dependent adjustment of accelerometer and gyroscope fusion weights. We investigate to what extent these limitations can be overcome by means of artificial neural networks and how much domain-specific optimization of the neural network model is required to outperform the conventional filter solution. A diverse set of motion recordings with a marker-based optical ground truth is used for performance evaluation and comparison. The proposed neural networks are found to outperform the conventional filter across all motions only if domain-specific optimizations are introduced. We conclude that they are a promising tool for inertial-sensor-based real-time attitude estimation, but both expert knowledge and rich datasets are required to achieve top performance.
AB - Inertial measurement units are commonly used to estimate the attitude of moving objects. Numerous nonlinear filter approaches have been proposed for solving the inherent sensor fusion problem. However, when a large range of different dynamic and static rotational and translational motions is considered, the attainable accuracy is limited by the need for situation-dependent adjustment of accelerometer and gyroscope fusion weights. We investigate to what extent these limitations can be overcome by means of artificial neural networks and how much domain-specific optimization of the neural network model is required to outperform the conventional filter solution. A diverse set of motion recordings with a marker-based optical ground truth is used for performance evaluation and comparison. The proposed neural networks are found to outperform the conventional filter across all motions only if domain-specific optimizations are introduced. We conclude that they are a promising tool for inertial-sensor-based real-time attitude estimation, but both expert knowledge and rich datasets are required to achieve top performance.
KW - Attitude determination
KW - Convolutional neural networks
KW - Inertial sensors
KW - Neural networks
KW - Nonlinear filters
KW - Performance evaluation
KW - Recurrent neural networks
KW - Sensor fusion
UR - http://www.scopus.com/inward/record.url?scp=85092720123&partnerID=8YFLogxK
U2 - 10.23919/FUSION45008.2020.9190634
DO - 10.23919/FUSION45008.2020.9190634
M3 - Conference contribution
AN - SCOPUS:85092720123
T3 - Proceedings of 2020 23rd International Conference on Information Fusion, FUSION 2020
BT - Proceedings of 2020 23rd International Conference on Information Fusion, FUSION 2020
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 23rd International Conference on Information Fusion, FUSION 2020
Y2 - 6 July 2020 through 9 July 2020
ER -