Zicheng Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zicheng Wang has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Hua Meng
School of Mathematics, Southwest Jiaotong University, Chengdu, Sichuan, China
3 shared publications - Meng Yang
School of Mathematics, Southwest Jiaotong University, China
2 shared publications Qing Xie
School of Mathematics, Southwest Jiaotong University, Chengdu, China
1 shared publicationShucong Tian
1 shared publication- Wei Xie
Department of Mechanical and Industrial Engineering Northeastern University Boston Massachusetts USA
1 shared publication Xi Chen
1 shared publication
Latest Publications
- 25 August 2026
- Published by Elsevier in Physical Communication
- Vol. 80
Maximum information gain reinforcement learning based on the variational information bottleneck
- 25 August 2026
- Published by Elsevier in Physical Communication
- Vol. 80
Maximum information gain reinforcement learning based on the variational information bottleneck
- 5 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (12) , 22178-22189
A Radar Signal Deinterleaving Method Based on Mixture Distribution Test With a Sequential Two-Module Neural Network Architecture
Radar signal deinterleaving (RSD) through pulse description word (PDW) clustering is a crucial pre-step for electronic reconnaissance. Although clustering algorithms are usually universal, existing PDW clustering algorithms cannot achieve a satisfying performance in a dense and complex electromagnetic environment. This paper presents a hierarchical clustering method (MDT-HC) based on the mixture distribution test, which combines learning the simulated PDW’s distribution as prior. The algorithm starts by regressing the marginal distribution of PDW attributes to determine the cluster number and the sub-distribution’s parameter information, such as the mean value, semi-range, and relative proportion. Then, the data are divided into over-lapping and non-overlapping areas referenced by these parameters, and the points in the overlapping area are iteratively classified to both sides using k-nearest neighbor (KNN) with non-overlapping points’ pure statistical features serving as a reference. The clustered result can be further sent into a trained sequential network for radiation source identification. The proposed algorithm’s time complexity and purity’s upper bound are analyzed theoretically from the perspective of geometric distribution properties. Its effectiveness and robustness have been validated through the measured data in the experiments. Besides, it has achieved the top 1% in the Big Data Challenge competition.
- 5 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 25 (12) , 22178-22189
A Radar Signal Deinterleaving Method Based on Mixture Distribution Test With a Sequential Two-Module Neural Network Architecture
Radar signal deinterleaving (RSD) through pulse description word (PDW) clustering is a crucial pre-step for electronic reconnaissance. Although clustering algorithms are usually universal, existing PDW clustering algorithms cannot achieve a satisfying performance in a dense and complex electromagnetic environment. This paper presents a hierarchical clustering method (MDT-HC) based on the mixture distribution test, which combines learning the simulated PDW’s distribution as prior. The algorithm starts by regressing the marginal distribution of PDW attributes to determine the cluster number and the sub-distribution’s parameter information, such as the mean value, semi-range, and relative proportion. Then, the data are divided into over-lapping and non-overlapping areas referenced by these parameters, and the points in the overlapping area are iteratively classified to both sides using k-nearest neighbor (KNN) with non-overlapping points’ pure statistical features serving as a reference. The clustered result can be further sent into a trained sequential network for radiation source identification. The proposed algorithm’s time complexity and purity’s upper bound are analyzed theoretically from the perspective of geometric distribution properties. Its effectiveness and robustness have been validated through the measured data in the experiments. Besides, it has achieved the top 1% in the Big Data Challenge competition.
- 29 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 60 (4) , 4463-4478
Reinforcement Learning-Based MIMO Radar Multitarget Detection Assisted by Bayesian Inference
Reinforcement Learning (RL) has been used to implement the perception-action cycle of multi-input-multi-output (MIMO) cognitive radar (CR). This allows for the adaptive optimization of the radar's beampattern, which is guided by information from echoes and an appropriate reward signal. However, the present approaches rely on a greedy DOA estimation to select candidate angles, which means that decisions are primarily based on detection results from the last pulse. If the system misses a target, it can be time-consuming to recapture it, and this can limit the detection performance. To address this issue, we develop a data-driven method for executing DOA estimation by utilizing the Bayesian inference to evaluate the likelihood of an angle containing a target based on historical detection information. Furthermore, to accommodate dynamic scenarios, a decay method for historical experience is proposed, allowing the system to adapt to environmental changes dynamically. The simulation results show that the RL-based MIMO radar with our refined DOA estimation module outperforms existing RL-based detector, providing the SOTA detection performance by focusing more frequently on essential angles, even in small-scale system setups.
- 29 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 60 (4) , 4463-4478
Reinforcement Learning-Based MIMO Radar Multitarget Detection Assisted by Bayesian Inference
Reinforcement Learning (RL) has been used to implement the perception-action cycle of multi-input-multi-output (MIMO) cognitive radar (CR). This allows for the adaptive optimization of the radar's beampattern, which is guided by information from echoes and an appropriate reward signal. However, the present approaches rely on a greedy DOA estimation to select candidate angles, which means that decisions are primarily based on detection results from the last pulse. If the system misses a target, it can be time-consuming to recapture it, and this can limit the detection performance. To address this issue, we develop a data-driven method for executing DOA estimation by utilizing the Bayesian inference to evaluate the likelihood of an angle containing a target based on historical detection information. Furthermore, to accommodate dynamic scenarios, a decay method for historical experience is proposed, allowing the system to adapt to environmental changes dynamically. The simulation results show that the RL-based MIMO radar with our refined DOA estimation module outperforms existing RL-based detector, providing the SOTA detection performance by focusing more frequently on essential angles, even in small-scale system setups.