Minti Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Minti Liu has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Cao Zeng
Xidian University
4 shared publications - Haihong Tao
Xidian University
2 shared publications - Jun Li
National Key Laboratory of Radar Signal Processing, Xidian University, Xian, China
2 shared publications - Guisheng Liao
Xidian University
1 shared publication Jiaqi Zhang
National Key Laboratory of Radar Signal Processing and the School of Electronic Engineering, Xidian University, Xian, Shaanxi, China
1 shared publication- Qinghua Guo
University of Wollongong
1 shared publication
Latest Publications
- 9 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (2) , 2471-2481
Informer-Based Joint Tracking and Trajectory Prediction of Maneuvering Radar Targets
Tracking radar targets over extended time horizons remains challenging due to highly nonlinear and maneuverable motion patterns that lead to cumulative prediction errors. This work addresses the joint estimation of current states and future trajectory prediction for radar targets under incomplete and noisy measurements. We develop a rigorous error-theoretic framework that formally analyzes error propagation mechanisms in sequential prediction, demonstrating that incremental state prediction effectively suppresses systematic bias and cumulative error growth compared to absolute prediction. Guided by these theoretical insights, we propose principled design strategies including feature augmentation with Doppler and kinematic information and increased network capacity. We instantiate this theory-guided approach within the Informer architecture, leveraging its efficient sequence-to-sequence capability to mitigate error accumulation while achieving O(LlogL) complexity, and propose this Informer-based radar maneuvering target tracking and Predicting framework (InforMTP). A synthetic dataset with randomized motion patterns and multi-mode dynamics is constructed to validate the framework. Experimental results demonstrate that the proposed method consistently achieves tighter error bounds and substantially outperforms the model-based and hybrid baselines in both prediction accuracy and convergence speed, particularly under complex maneuvering scenarios.
- 9 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (2) , 2471-2481
Informer-Based Joint Tracking and Trajectory Prediction of Maneuvering Radar Targets
Tracking radar targets over extended time horizons remains challenging due to highly nonlinear and maneuverable motion patterns that lead to cumulative prediction errors. This work addresses the joint estimation of current states and future trajectory prediction for radar targets under incomplete and noisy measurements. We develop a rigorous error-theoretic framework that formally analyzes error propagation mechanisms in sequential prediction, demonstrating that incremental state prediction effectively suppresses systematic bias and cumulative error growth compared to absolute prediction. Guided by these theoretical insights, we propose principled design strategies including feature augmentation with Doppler and kinematic information and increased network capacity. We instantiate this theory-guided approach within the Informer architecture, leveraging its efficient sequence-to-sequence capability to mitigate error accumulation while achieving O(LlogL) complexity, and propose this Informer-based radar maneuvering target tracking and Predicting framework (InforMTP). A synthetic dataset with randomized motion patterns and multi-mode dynamics is constructed to validate the framework. Experimental results demonstrate that the proposed method consistently achieves tighter error bounds and substantially outperforms the model-based and hybrid baselines in both prediction accuracy and convergence speed, particularly under complex maneuvering scenarios.
- 22 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (2) , 2998-3013
Joint One-Hop Flooding Communication and Event-Triggered Distributed Trajectory Fusion in Spatiotemporal Constraint Sensor Networks
In this work, we investigate the challenging problem of multitarget relay tracking with random appearance and disappearance in spatiotemporal sensor networks. In this scenario, sensor networks can be coordinated in a decentralized manner. In response, we propose a novel distributed trajectory fusion (DTF) method. Each local sensor, operating without prior knowledge of the network topology, performs different tasks based on target-oriented local trajectory estimation in three modes: source transmission mode (STM), passive receiving mode (PRM), and intermediate relay mode (IRM). Specifically, STM focuses on computing one-hop direct neighbors, PRM deals with the trajectory fusion, and IRM involves efficient communication through joint one-hop flooding and an event-triggered scheme, ensuring robust trajectory fusion. Our method incorporates Gaussian mixture approximated trajectory Poisson multi-Bernoulli filter with spatiotemporal constraints (GM-TPMB-STC) to accurately capture expected target movements as they enter the area of interest. Simulation results demonstrate the superiority of our proposed method over existing distributed target state and density fusion methods in both accuracy and computational efficiency, offering a promising solution to multitarget tracking in complex environments.
- 22 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (2) , 2998-3013
Joint One-Hop Flooding Communication and Event-Triggered Distributed Trajectory Fusion in Spatiotemporal Constraint Sensor Networks
In this work, we investigate the challenging problem of multitarget relay tracking with random appearance and disappearance in spatiotemporal sensor networks. In this scenario, sensor networks can be coordinated in a decentralized manner. In response, we propose a novel distributed trajectory fusion (DTF) method. Each local sensor, operating without prior knowledge of the network topology, performs different tasks based on target-oriented local trajectory estimation in three modes: source transmission mode (STM), passive receiving mode (PRM), and intermediate relay mode (IRM). Specifically, STM focuses on computing one-hop direct neighbors, PRM deals with the trajectory fusion, and IRM involves efficient communication through joint one-hop flooding and an event-triggered scheme, ensuring robust trajectory fusion. Our method incorporates Gaussian mixture approximated trajectory Poisson multi-Bernoulli filter with spatiotemporal constraints (GM-TPMB-STC) to accurately capture expected target movements as they enter the area of interest. Simulation results demonstrate the superiority of our proposed method over existing distributed target state and density fusion methods in both accuracy and computational efficiency, offering a promising solution to multitarget tracking in complex environments.
- 3 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (4) , 4722-4734
An Adaptive Track Initiation Method Adopting Multiscan Spatiotemporal Information
Trajectory initiation of dense newborn targets is an important research topic in multi-target tracking, however, few reports concerning the related studies can be found. In response to the problem of accurately obtaining spatiotemporal information of densely populated newborn targets, we propose a novel adaptive track initiation method adopting multi-scan spatiotemporal information. Firstly, by combining the spatial information measured from multiple scans, an adaptive track initiation method is proposed to obtain the measurement sequence of newborn targets. Secondly, by establishing an adaptive newborn density model, the newborn state estimation of dense targets is obtained from the measurement sequence of newborn targets. Finally, the state estimation of the targets is input to the labelled random finite set filter to obtain the trajectory information. By quantifying the posterior density error of the labelled random finite set, the truncation error of the newborn state estimation is presented. The error analysis suggests that our proposed method can achieve track initiation with small boundary L1 error. The simulation results verify that the proposed method can effectively cope with the situation of dense newborn targets. Compared with the existing adaptive birth algorithms, our proposal can realize more accurate track initiation when newborn targets occur simultaneously.
- 3 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 24 (4) , 4722-4734
An Adaptive Track Initiation Method Adopting Multiscan Spatiotemporal Information
Trajectory initiation of dense newborn targets is an important research topic in multi-target tracking, however, few reports concerning the related studies can be found. In response to the problem of accurately obtaining spatiotemporal information of densely populated newborn targets, we propose a novel adaptive track initiation method adopting multi-scan spatiotemporal information. Firstly, by combining the spatial information measured from multiple scans, an adaptive track initiation method is proposed to obtain the measurement sequence of newborn targets. Secondly, by establishing an adaptive newborn density model, the newborn state estimation of dense targets is obtained from the measurement sequence of newborn targets. Finally, the state estimation of the targets is input to the labelled random finite set filter to obtain the trajectory information. By quantifying the posterior density error of the labelled random finite set, the truncation error of the newborn state estimation is presented. The error analysis suggests that our proposed method can achieve track initiation with small boundary L1 error. The simulation results verify that the proposed method can effectively cope with the situation of dense newborn targets. Compared with the existing adaptive birth algorithms, our proposal can realize more accurate track initiation when newborn targets occur simultaneously.
- 21 November 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 23 (1) , 470-483
Robust TPMB Filtering Using Sensors With Limited Sensing Range Under Nonuniform Clutter Background
This study proposes a novel solution using sensors with limited sensing range to estimate the random appearance or disappearance of multi-object trajectories with unknown object detection profile under nonuniform clutter background. Specifically, assuming that no objects occur within the sensors’ observable area, the unknown clutter intensity involving the average false alarms per scan and nonuniform clutter density is first estimated adopting the joint combination of maximum likelihood (ML) and model-based clustering methods. Then, the unknown object detection profile can be calculated via marginalizing the fluctuating SNR in a random manner. Further, the trajectory Poisson multi-Bernoulli (TPMB) filter is utilized to estimate trajectories from first principle, after which the estimated parameters of clutter intensity and the detection probability are fed to the TPMB filtering, thereby improving the performance on completeness and robustness of estimation trajectories. Simulation and experimental results demonstrated that our proposed solution exhibits excellent tracking accuracy comparing with state-of-the-art solutions involving robust multi-sensor generalized labeled multi-Bernoulli (R-MS-GLMB) and dynamic parameter generalized labeled multi-Bernoulli (DP-GLMB) filters.
- 21 November 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 23 (1) , 470-483
Robust TPMB Filtering Using Sensors With Limited Sensing Range Under Nonuniform Clutter Background
This study proposes a novel solution using sensors with limited sensing range to estimate the random appearance or disappearance of multi-object trajectories with unknown object detection profile under nonuniform clutter background. Specifically, assuming that no objects occur within the sensors’ observable area, the unknown clutter intensity involving the average false alarms per scan and nonuniform clutter density is first estimated adopting the joint combination of maximum likelihood (ML) and model-based clustering methods. Then, the unknown object detection profile can be calculated via marginalizing the fluctuating SNR in a random manner. Further, the trajectory Poisson multi-Bernoulli (TPMB) filter is utilized to estimate trajectories from first principle, after which the estimated parameters of clutter intensity and the detection probability are fed to the TPMB filtering, thereby improving the performance on completeness and robustness of estimation trajectories. Simulation and experimental results demonstrated that our proposed solution exhibits excellent tracking accuracy comparing with state-of-the-art solutions involving robust multi-sensor generalized labeled multi-Bernoulli (R-MS-GLMB) and dynamic parameter generalized labeled multi-Bernoulli (DP-GLMB) filters.