TY - JOUR
T1 - Improved probability hypothesis density filter for inaccurate model and heavy-tailed noise
AU - Zhang, Yanzhuo
AU - Yan, Liping
AU - Zhou, Yuqin
AU - Xia, Yuanqing
N1 - Publisher Copyright:
© 2026 Elsevier Masson SAS.
PY - 2027/1
Y1 - 2027/1
N2 - In the field of multi-target tracking (MTT) based on the random finite set, the Gaussian mixture probability hypothesis density (GM-PHD) filter stands out for its outstanding target tracking performance, such as its ability to avoid explicit measurement-to-target data association and its ease of implementation. However, effective tracking with the GM-PHD filter relies on the assumption of an accurate motion model with Gaussian noise. From a modeling perspective, a key challenge in practical tracking systems is model inaccuracy, which refers to the inability to obtain prior knowledge of the target’s true dynamic model, inevitably leading to discrepancies between the filter’s assumed model and the actual one, thereby causing estimation bias. Generally, the dynamic model of target includes the motion model and the noise model. Therefore, addressing tracking problems under model inaccuracy requires simultaneous consideration of inaccuracies in both the target motion model and the noise model. To solve these problems, this paper proposes a Student’s t mixture-based strong tracking PHD (StM-STPHD) filter. Firstly, to address the issue of target motion model inaccuracy, an improved GM-PHD filter is presented by introducing the strong tracking filter into the prediction and update processes of the GM-PHD filter, namely the ST-GMPHD filter. Secondly, to address the simultaneous presence of an inaccurate motion model and heavy-tailed noise, the StM-STPHD filter is designed by integrating Student’s t mixture components into the proposed ST-GMPHD filter. Finally, the effectiveness of the ST-GMPHD filter and the StM-STPHD filter is verified through multiple groups of simulation experiments.
AB - In the field of multi-target tracking (MTT) based on the random finite set, the Gaussian mixture probability hypothesis density (GM-PHD) filter stands out for its outstanding target tracking performance, such as its ability to avoid explicit measurement-to-target data association and its ease of implementation. However, effective tracking with the GM-PHD filter relies on the assumption of an accurate motion model with Gaussian noise. From a modeling perspective, a key challenge in practical tracking systems is model inaccuracy, which refers to the inability to obtain prior knowledge of the target’s true dynamic model, inevitably leading to discrepancies between the filter’s assumed model and the actual one, thereby causing estimation bias. Generally, the dynamic model of target includes the motion model and the noise model. Therefore, addressing tracking problems under model inaccuracy requires simultaneous consideration of inaccuracies in both the target motion model and the noise model. To solve these problems, this paper proposes a Student’s t mixture-based strong tracking PHD (StM-STPHD) filter. Firstly, to address the issue of target motion model inaccuracy, an improved GM-PHD filter is presented by introducing the strong tracking filter into the prediction and update processes of the GM-PHD filter, namely the ST-GMPHD filter. Secondly, to address the simultaneous presence of an inaccurate motion model and heavy-tailed noise, the StM-STPHD filter is designed by integrating Student’s t mixture components into the proposed ST-GMPHD filter. Finally, the effectiveness of the ST-GMPHD filter and the StM-STPHD filter is verified through multiple groups of simulation experiments.
KW - Multi-target tracking
KW - Probability hypothesis density filter
KW - Random finite set
KW - Strong tracking filter
KW - Student’ s t distribution
UR - https://www.scopus.com/pages/publications/105054205359
U2 - 10.1016/j.ast.2026.114062
DO - 10.1016/j.ast.2026.114062
M3 - Article
AN - SCOPUS:105054205359
SN - 1270-9638
VL - 180
JO - Aerospace Science and Technology
JF - Aerospace Science and Technology
M1 - 114062
ER -