Fengzhong Qu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Fengzhong Qu has a total of 353 co-authors. Here are the top co-authors by number of shared publications:
- Yan Wei
Zhejiang University
34 shared publications - Xingbin Tu
Key Laboratory of Ocean Observation-Imaging Testbed of Zhejiang Province, Zhejiang University, Zhoushan, China
24 shared publications Liuqing Yang
Ocean College, Zhejiang University, Hangzhou 310058, China
20 shared publications- Jiang Zhu
Zhejiang University
11 shared publications - Xiaoxiao Zhuo
Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences
11 shared publications Jun Han
10 shared publications
Latest Publications
- Open Access
- 3 November 2026
- Published by MDPI AG in Journal of Marine Science and Engineering
- Vol. 14 (21) , 2068
Geometry-Constrained Joint Multipath Exploitation for Single-Transceiver Underwater Acoustic Localization
Boundary reflections are usually treated as interference in underwater acoustic localization, yet their position-dependent delays can form a virtual aperture for a single transceiver. This paper develops a geometry-constrained method for exploiting that information without calibrated reflection amplitudes or phases. Reciprocal direct, surface, and bottom echoes are described by image-source delays, while their complex gains remain nuisance parameters. Variable projection yields a position-dependent joint multipath exploitation (JME) objective. The equivalent Fisher information relates local accuracy to bandwidth, path energy, resolvability, and delay-gradient geometry. Grid-search initialization (GSI) and delay-geometric initialization (DGI) feed a common safeguarded Newton-like refinement (JME-NR). Simulations include a coherent matched-field processing baseline and environmental mismatch tests. Among the JME routes, GSI is more reliable at very low signal-to-noise ratio (SNR), whereas DGI becomes more effective with reliable delay extraction and path association. For a fixed target at , GSI-JME-NR achieves an all-trial root-mean-square error (RMSE) of , close to the exact position error bound of . In finite-tank tests with a separated transmitter and receiver, three two-dimensional configurations with 20 measurements each yield refined RMSEs of – relative to the target’s geometric center. Three-dimensional tests show that improved waveform fitting does not necessarily reduce the error relative to this center.- 3 November 2026
- Published by MDPI AG in Journal of Marine Science and Engineering
- Vol. 14 (21) , 2068
Open AccessGeometry-Constrained Joint Multipath Exploitation for Single-Transceiver Underwater Acoustic Localization
Boundary reflections are usually treated as interference in underwater acoustic localization, yet their position-dependent delays can form a virtual aperture for a single transceiver. This paper develops a geometry-constrained method for exploiting that information without calibrated reflection amplitudes or phases. Reciprocal direct, surface, and bottom echoes are described by image-source delays, while their complex gains remain nuisance parameters. Variable projection yields a position-dependent joint multipath exploitation (JME) objective. The equivalent Fisher information relates local accuracy to bandwidth, path energy, resolvability, and delay-gradient geometry. Grid-search initialization (GSI) and delay-geometric initialization (DGI) feed a common safeguarded Newton-like refinement (JME-NR). Simulations include a coherent matched-field processing baseline and environmental mismatch tests. Among the JME routes, GSI is more reliable at very low signal-to-noise ratio (SNR), whereas DGI becomes more effective with reliable delay extraction and path association. For a fixed target at , GSI-JME-NR achieves an all-trial root-mean-square error (RMSE) of , close to the exact position error bound of . In finite-tank tests with a separated transmitter and receiver, three two-dimensional configurations with 20 measurements each yield refined RMSEs of – relative to the target’s geometric center. Three-dimensional tests show that improved waveform fitting does not necessarily reduce the error relative to this center. - 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (21) , 32029-32043
The Development and Application of Laser-Induced Acoustics in Cross-Air-Sea Interface Communication: A Review
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (21) , 32029-32043
The Development and Application of Laser-Induced Acoustics in Cross-Air-Sea Interface Communication: A Review
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (21) , 51032-51049
SRL-NTO: Shared Reinforcement Learning-Based Network Topology Optimization for UASNs
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (21) , 51032-51049
SRL-NTO: Shared Reinforcement Learning-Based Network Topology Optimization for UASNs
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (11) , 21426-21441
A Low-Complexity Transmission Constraint for Collision-Free Single-Hop Broadcasting in Underwater Acoustic Networks
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (11) , 21426-21441
A Low-Complexity Transmission Constraint for Collision-Free Single-Hop Broadcasting in Underwater Acoustic Networks
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Oceanic Engineering
- Vol. 51 (3) , 2478-2494
An Effective Receptive Field-Guided Parallel Resolution ConvNet for Underwater Salient Object Detection
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Oceanic Engineering
- Vol. 51 (3) , 2478-2494
An Effective Receptive Field-Guided Parallel Resolution ConvNet for Underwater Salient Object Detection
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Oceanic Engineering
- Vol. 51 (3) , 1662-1675
A Phase Noise Reduction Method for Reliable Cross-Medium Communication
Cross-medium communication between underwater and aerial environments is crucial for marine exploration, environmental monitoring, and surveillance. A promising method uses millimeter-wave radar to detect microscale water surface vibrations induced by underwater acoustic sources, but its performance is limited by radar phase noise. To address this, we present a novel application of data-driven differential phase processing tailored to this cross-medium scenario. Our approach integrates adaptive truncation to mitigate hardware-induced nonlinearity and differential processing to suppress environmental noise, thereby enhancing the signal-to-noise ratio and enabling accurate data recovery. The method has been experimentally validated in both calm water and artificially generated wavy conditions, demonstrating its robustness and effectiveness in improving communication performance. This work establishes a robust processing framework for cross-medium communication in varied environments.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Oceanic Engineering
- Vol. 51 (3) , 1662-1675
A Phase Noise Reduction Method for Reliable Cross-Medium Communication
Cross-medium communication between underwater and aerial environments is crucial for marine exploration, environmental monitoring, and surveillance. A promising method uses millimeter-wave radar to detect microscale water surface vibrations induced by underwater acoustic sources, but its performance is limited by radar phase noise. To address this, we present a novel application of data-driven differential phase processing tailored to this cross-medium scenario. Our approach integrates adaptive truncation to mitigate hardware-induced nonlinearity and differential processing to suppress environmental noise, thereby enhancing the signal-to-noise ratio and enabling accurate data recovery. The method has been experimentally validated in both calm water and artificially generated wavy conditions, demonstrating its robustness and effectiveness in improving communication performance. This work establishes a robust processing framework for cross-medium communication in varied environments.
- 20 May 2026
- Published by Elsevier in Marine Pollution Bulletin
- Vol. 231, 119872
Spatiotemporal predictive modeling for coastal phytoplankton based on the Tapnet model: A case study in Zhejiang, China
- 20 May 2026
- Published by Elsevier in Marine Pollution Bulletin
- Vol. 231, 119872
Spatiotemporal predictive modeling for coastal phytoplankton based on the Tapnet model: A case study in Zhejiang, China
- 22 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (14) , 30733-30747
MAC Protocol for ISAC in Underwater Acoustic Networks With Dynamic Target Monitoring–Localization–Tracking
This paper proposes a multiple access control (MAC) protocol for integrated sensing and communication in underwater acoustic networks (ISAC-UANs) with dynamic target monitoring, localization, and tracking, where some ISAC nodes are deployed to collaboratively communicate and detect targets via integrated underwater acoustic signals. The proposed integrated underwater sensing and communication (USAC)-MAC protocol focuses on addressing the challenges of packet collisions and resource contention in the dual-function network. Specifically, the protocol is divided into three stages: target monitoring, target localization, and the target tracking phase. The target monitoring phase is carried out to initialize the network topology and search for detection targets. The conflict-free concurrent transmission constraints are established based on topological information to enable conflict-free and efficient data transmission. After detecting the target, the protocol transfers into the target localization phase, which localizes the target applying the time of arrival (TOA) method. We model the optimization problem to minimize the Cramér-Rao lower bound (CRLB) and employ the particle swarm optimization (PSO) algorithm to solve it under the constraint of the geometric position of nodes. During the target tracking phase, we enhance target detection accuracy by dynamically selecting detection node clusters based on predicted target trajectories. Concurrently, we schedule packet transmission times to minimize the communication frame length. Thereby, a joint optimization problem for the CRLB of localization and the frame length of communication is established and addressed by the simulated annealing (SA) algorithm, ensuring detection precision while reducing communication latency. Simulation results demonstrate that the proposed protocol exhibits superior performance with higher throughput, shorter frame length, shorter information update times, and more precise detection accuracy.
- 22 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (14) , 30733-30747
MAC Protocol for ISAC in Underwater Acoustic Networks With Dynamic Target Monitoring–Localization–Tracking
This paper proposes a multiple access control (MAC) protocol for integrated sensing and communication in underwater acoustic networks (ISAC-UANs) with dynamic target monitoring, localization, and tracking, where some ISAC nodes are deployed to collaboratively communicate and detect targets via integrated underwater acoustic signals. The proposed integrated underwater sensing and communication (USAC)-MAC protocol focuses on addressing the challenges of packet collisions and resource contention in the dual-function network. Specifically, the protocol is divided into three stages: target monitoring, target localization, and the target tracking phase. The target monitoring phase is carried out to initialize the network topology and search for detection targets. The conflict-free concurrent transmission constraints are established based on topological information to enable conflict-free and efficient data transmission. After detecting the target, the protocol transfers into the target localization phase, which localizes the target applying the time of arrival (TOA) method. We model the optimization problem to minimize the Cramér-Rao lower bound (CRLB) and employ the particle swarm optimization (PSO) algorithm to solve it under the constraint of the geometric position of nodes. During the target tracking phase, we enhance target detection accuracy by dynamically selecting detection node clusters based on predicted target trajectories. Concurrently, we schedule packet transmission times to minimize the communication frame length. Thereby, a joint optimization problem for the CRLB of localization and the frame length of communication is established and addressed by the simulated annealing (SA) algorithm, ensuring detection precision while reducing communication latency. Simulation results demonstrate that the proposed protocol exhibits superior performance with higher throughput, shorter frame length, shorter information update times, and more precise detection accuracy.
- 3 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (10) , 22440-22461
RLMA-UNTO: Reinforcement Learning and Metaheuristic Algorithms-Based Intelligent Topology Optimization for UASNs
Multinode underwater acoustic sensor networks (UASNs) are critical for applications such as environmental monitoring and target detection. However, they face unique challenges, including long propagation delays, narrow bandwidth, and the complex marine environment. This makes topology optimization for achieving high coverage, low delay, and high throughput difficult, especially under high-load application scenarios. To address these difficulties, we propose a unified framework, termed RLMA-UNTO, which integrates reinforcement learning (RL) with metaheuristic algorithms for multiobjective topology optimization in UASNs. The framework synergistically combines the global exploration strengths of population-based metaheuristics with the local refinement capabilities of an actor–critic RL module. Simulations under high-load conditions for network sizes of 15–70 nodes show that RLMA-UNTO consistently outperforms its traditional metaheuristic variants, as well as benchmark optimizers. Specifically, averaged over all network sizes from 15 to 70 nodes, the RLMA-UNTO-based optimizers achieve consistently higher coverage than their metaheuristic counterparts. The average network throughput is increased by about 20 to 24 bits per second (bps). At the same time, the average end-to-end delay is reduced by 43 to 56 seconds (s), corresponding to a 23%–25% reduction. Similar performance margins are also observed when comparing these RL-enhanced schemes with the benchmark optimizers.
- 3 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (10) , 22440-22461
RLMA-UNTO: Reinforcement Learning and Metaheuristic Algorithms-Based Intelligent Topology Optimization for UASNs
Multinode underwater acoustic sensor networks (UASNs) are critical for applications such as environmental monitoring and target detection. However, they face unique challenges, including long propagation delays, narrow bandwidth, and the complex marine environment. This makes topology optimization for achieving high coverage, low delay, and high throughput difficult, especially under high-load application scenarios. To address these difficulties, we propose a unified framework, termed RLMA-UNTO, which integrates reinforcement learning (RL) with metaheuristic algorithms for multiobjective topology optimization in UASNs. The framework synergistically combines the global exploration strengths of population-based metaheuristics with the local refinement capabilities of an actor–critic RL module. Simulations under high-load conditions for network sizes of 15–70 nodes show that RLMA-UNTO consistently outperforms its traditional metaheuristic variants, as well as benchmark optimizers. Specifically, averaged over all network sizes from 15 to 70 nodes, the RLMA-UNTO-based optimizers achieve consistently higher coverage than their metaheuristic counterparts. The average network throughput is increased by about 20 to 24 bits per second (bps). At the same time, the average end-to-end delay is reduced by 43 to 56 seconds (s), corresponding to a 23%–25% reduction. Similar performance margins are also observed when comparing these RL-enhanced schemes with the benchmark optimizers.