TY - GEN
T1 - Performance Evaluation of Simultaneous Sensor Registration and Object Tracking Algorithm
AU - MacDonald, Sofie
AU - Proudler, Ian
AU - Davies, Michael E.
AU - Hopgood, James R.
N1 - Funding Information:
S. Macdonald, M. E. Davies, and J. R. Hopgood are with the School of Engineering, University of Edinburgh, Edinburgh. E-mail: (s1229110; Mike.Davies; James.Hopgood)@ed.ac.uk. I. Proudler is with the Department of Electronic & Electrical Engineering, University of Strathclyde, Glasgow. E-mail: [email protected]. This work was supported by EPSRC Grant EP/S000631/1; the MOD University Defence Research Collaboration in Signal Processing; and the Look Out (AEW) themed competition, run by the Defence and Security Accelerator on behalf of The Royal Navy and DIU; and Leonardo MW Ltd. https://uk.leonardocompany.com/en/home
Publisher Copyright:
© 2022 IEEE.
PY - 2022/10/13
Y1 - 2022/10/13
N2 - Reliable object tracking with multiple sensors requires that sensors are registered correctly with respect to each other. When an environment is Global Navigation Satellite System (GNSS) denied or limited – such as underwater, or in hostile regions – this task is more challenging. This paper performs uncertainty quantification on a simultaneous tracking and registration algorithm for sensor networks that does not require access to a GNSS. The method uses a particle filter combined with a bank of augmented state extended Kalman filters (EKFs). The particles represent hypotheses of registration errors between sensors, with associated weights. The EKFs are responsible for the tracking procedure and for contributing to particle state and weight updates. This is achieved through the evaluation of a likelihood. Registration errors in this paper are spatial, orientation, and temporal biases: seven distinct sensor errors are estimated alongside the tracking procedure. Monte Carlo trials are conducted for the uncertainty quantification. Since performance of particle filters is dependent on initialisation, a comparison is made between more and less favourable particle (hypothesis) initialisation. The results demonstrate the importance of initialisation, and the method is shown to perform well in tracking a fast (marginally sub-sonic) object following a bow-like trajectory (mimicking a representative scenario). Final results show the algorithm is capable of achieving angular bias estimation error of 0.0034 degrees, temporal bias estimation error of 0.0067s, and spatial error of 0.021m.
AB - Reliable object tracking with multiple sensors requires that sensors are registered correctly with respect to each other. When an environment is Global Navigation Satellite System (GNSS) denied or limited – such as underwater, or in hostile regions – this task is more challenging. This paper performs uncertainty quantification on a simultaneous tracking and registration algorithm for sensor networks that does not require access to a GNSS. The method uses a particle filter combined with a bank of augmented state extended Kalman filters (EKFs). The particles represent hypotheses of registration errors between sensors, with associated weights. The EKFs are responsible for the tracking procedure and for contributing to particle state and weight updates. This is achieved through the evaluation of a likelihood. Registration errors in this paper are spatial, orientation, and temporal biases: seven distinct sensor errors are estimated alongside the tracking procedure. Monte Carlo trials are conducted for the uncertainty quantification. Since performance of particle filters is dependent on initialisation, a comparison is made between more and less favourable particle (hypothesis) initialisation. The results demonstrate the importance of initialisation, and the method is shown to perform well in tracking a fast (marginally sub-sonic) object following a bow-like trajectory (mimicking a representative scenario). Final results show the algorithm is capable of achieving angular bias estimation error of 0.0034 degrees, temporal bias estimation error of 0.0067s, and spatial error of 0.021m.
U2 - 10.1109/MFI55806.2022.9913857
DO - 10.1109/MFI55806.2022.9913857
M3 - Conference contribution
SN - 978-1-6654-6027-9
BT - 2022 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI)
PB - Institute of Electrical and Electronics Engineers
T2 - 2022 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI)
Y2 - 20 September 2022 through 22 September 2022
ER -