Tracking multiple space objects using multiple surveillance sensors is a critical approach in many Space Situation Awareness(SSA) applications. In this process, the uncertainties of targets,dynamics, and observations ...Tracking multiple space objects using multiple surveillance sensors is a critical approach in many Space Situation Awareness(SSA) applications. In this process, the uncertainties of targets,dynamics, and observations are usually represented by the probability distributions. However, precise characterization of uncertainty becomes challenging due to imperfect knowledge about some key aspects, such as birth targets and sensor detection profiles. To overcome this challenge, this paper proposes a multi-sensor possibility PHD filter based on the theory of outer probability measures. An effective compensation method is introduced to tackle variations in the fields of view of SSA sensors or instances of missed detections, aiming to mitigate the inconsistency in localized information. The proposed method is adapted to centralized and distributed sensor networks, offering effective solutions for multi-sensor multi-target tracking. The major innovation of the proposed method compared with typical methods is the proper description of epistemic uncertainty, which yields more robust performance in the scenarios of lacking some information about the system.The effectiveness of the multi-sensor possibility PHD filter is demonstrated by a comparison with conventional methods in two simulated scenarios.展开更多
基金funded by the National Natural Science Foundation of China(No.12202049)the Beijing Institute of Technology Research Fund Program for Young Scholars,China.
文摘Tracking multiple space objects using multiple surveillance sensors is a critical approach in many Space Situation Awareness(SSA) applications. In this process, the uncertainties of targets,dynamics, and observations are usually represented by the probability distributions. However, precise characterization of uncertainty becomes challenging due to imperfect knowledge about some key aspects, such as birth targets and sensor detection profiles. To overcome this challenge, this paper proposes a multi-sensor possibility PHD filter based on the theory of outer probability measures. An effective compensation method is introduced to tackle variations in the fields of view of SSA sensors or instances of missed detections, aiming to mitigate the inconsistency in localized information. The proposed method is adapted to centralized and distributed sensor networks, offering effective solutions for multi-sensor multi-target tracking. The major innovation of the proposed method compared with typical methods is the proper description of epistemic uncertainty, which yields more robust performance in the scenarios of lacking some information about the system.The effectiveness of the multi-sensor possibility PHD filter is demonstrated by a comparison with conventional methods in two simulated scenarios.
