Xing Wei
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xing Wei has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Hua Yang
Macao Polytechnic University, Faculty of Applied Sciences, Macao 999078, P. R. China
3 shared publications - Dennis Wong
Macao Polytechnic University
2 shared publications - Bo Li
Guangxi Key Laboratory of Low-Altitude Unmanned Autonomous Systems, Guilin University of Aerospace Technology, Guilin, China and School of Computer Science and Engineering, Guilin University of Aerospace Technology, Guilin, China
1 shared publication - Jing Lu
School of Information Engineering, Southwest University of Science and Technology, Mianyang 621010, China
1 shared publication - Jingyi Yan
Guilin Institute of Information Technology, Business School, China
1 shared publication - Ning Li
Guilin University of Aerospace Technology, School of Computer Science and Engineering Guilin 541004, P. R. China
1 shared publication
Latest Publications
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7475-7480
Remote Sensing Image-Text Retrieval via Hierarchical and Targeting Semantic Alignment
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7475-7480
Open AccessRemote Sensing Image-Text Retrieval via Hierarchical and Targeting Semantic Alignment
- 30 October 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Pattern Recognition and Artificial Intelligence
Adaptive FANET Clustering Integrating PSO-Optimized DBSCAN for 3D Dynamic Topologies
In Flying Ad Hoc Networks (FANETs), building a stable and efficient communication backbone faces significant challenges due to the highly dynamic three-dimensional topology, rapid node mobility, and frequent link interruptions. To overcome the inherent limitations of density-based spatial clustering algorithms like Density-based Spatial Clustering of Applications with Noise (DBSCAN), which heavily rely on two static parameters, the neighborhood radius [Formula: see text] and the minimum number of points MinPts, PSO-DBSCAN proposes a novel adaptive clustering protocol that integrates the Particle Swarm Optimization (PSO) algorithm with DBSCAN. The proposed method consists of three stages: First, PSO dynamically optimizes DBSCAN parameters based on the current network topology before each clustering period. Second, DBSCAN is executed using the optimized parameters to form density-based clusters. Finally, a second round of PSO is run within each cluster to elect the optimal cluster head based on a multi-objective fitness function that comprehensively considers residual energy, mobility, node degree, and distance to the cluster center. Simulations conducted on the NS-3 platform demonstrate that PSO-DBSCAN offers significant advantages over benchmark algorithms such as Weighted Clustering Algorithm (WCA) and DBSCAN in terms of clustering quality, cluster stability, control overhead, energy efficiency, and network lifetime. More importantly, the proposed scheme exhibits excellent adaptability to sudden topological changes, providing a robust and scalable solution for dynamic three-dimensional FANET environments.
- 30 October 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Pattern Recognition and Artificial Intelligence
Adaptive FANET Clustering Integrating PSO-Optimized DBSCAN for 3D Dynamic Topologies
In Flying Ad Hoc Networks (FANETs), building a stable and efficient communication backbone faces significant challenges due to the highly dynamic three-dimensional topology, rapid node mobility, and frequent link interruptions. To overcome the inherent limitations of density-based spatial clustering algorithms like Density-based Spatial Clustering of Applications with Noise (DBSCAN), which heavily rely on two static parameters, the neighborhood radius [Formula: see text] and the minimum number of points MinPts, PSO-DBSCAN proposes a novel adaptive clustering protocol that integrates the Particle Swarm Optimization (PSO) algorithm with DBSCAN. The proposed method consists of three stages: First, PSO dynamically optimizes DBSCAN parameters based on the current network topology before each clustering period. Second, DBSCAN is executed using the optimized parameters to form density-based clusters. Finally, a second round of PSO is run within each cluster to elect the optimal cluster head based on a multi-objective fitness function that comprehensively considers residual energy, mobility, node degree, and distance to the cluster center. Simulations conducted on the NS-3 platform demonstrate that PSO-DBSCAN offers significant advantages over benchmark algorithms such as Weighted Clustering Algorithm (WCA) and DBSCAN in terms of clustering quality, cluster stability, control overhead, energy efficiency, and network lifetime. More importantly, the proposed scheme exhibits excellent adaptability to sudden topological changes, providing a robust and scalable solution for dynamic three-dimensional FANET environments.
- 6 July 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Pattern Recognition and Artificial Intelligence
Topology-Aware Genetic Clustering with Elite Opposition-Based Learning for High-Mobility FANETs
To address the frequent cluster structure reconfiguration caused by the high mobility and dynamic topological changes in Flying Ad Hoc Networks (FANETs), this paper proposes an optimal Cluster Head (CH) selection mechanism based on an adaptive genetic clustering algorithm (AGCA). Traditional clustering methods often struggle with multi-objective conflicts and fixed-length encoding limitations in highly dynamic environments. The proposed AGCA dynamically characterizes the cluster topology using a variable-length genetic encoding scheme to adaptively respond to network scale variations. Furthermore, a comprehensive multi-objective fitness function is constructed to jointly optimize energy balance, communication distance, and topological stability. To enhance cluster quality and accelerate convergence, an Elite Opposition-Based Learning (EOL) strategy and a chaotic initialization method are integrated into the algorithm. Extensive simulation experiments demonstrate that, compared with the classical Weighted Clustering Algorithm (WCA), the proposed AGCA achieves superior performance in highly dynamic FANET environments. Specifically, the ACGA extends the average cluster head lifetime by 22%, reduces end-to-end delay by 23%, increases network throughput by 18%, and decreases control overhead by 30%, thereby verifying its scientific validity and robustness.
- 6 July 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Pattern Recognition and Artificial Intelligence
Topology-Aware Genetic Clustering with Elite Opposition-Based Learning for High-Mobility FANETs
To address the frequent cluster structure reconfiguration caused by the high mobility and dynamic topological changes in Flying Ad Hoc Networks (FANETs), this paper proposes an optimal Cluster Head (CH) selection mechanism based on an adaptive genetic clustering algorithm (AGCA). Traditional clustering methods often struggle with multi-objective conflicts and fixed-length encoding limitations in highly dynamic environments. The proposed AGCA dynamically characterizes the cluster topology using a variable-length genetic encoding scheme to adaptively respond to network scale variations. Furthermore, a comprehensive multi-objective fitness function is constructed to jointly optimize energy balance, communication distance, and topological stability. To enhance cluster quality and accelerate convergence, an Elite Opposition-Based Learning (EOL) strategy and a chaotic initialization method are integrated into the algorithm. Extensive simulation experiments demonstrate that, compared with the classical Weighted Clustering Algorithm (WCA), the proposed AGCA achieves superior performance in highly dynamic FANET environments. Specifically, the ACGA extends the average cluster head lifetime by 22%, reduces end-to-end delay by 23%, increases network throughput by 18%, and decreases control overhead by 30%, thereby verifying its scientific validity and robustness.
- 6 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (1) , 353-360
Continuous Hopfield Neural Network (CHNN)-Based Clustering Optimization Scheme for FANETs
With the rapid development of UAV technology, Flying Ad hoc Network (FANET) has become a core network architecture in military, disaster relief, and other fields due to its potential for efficient communication and collaboration in dynamic environments. However, existing clustering algorithms exhibit significant drawbacks in dynamic adaptability, parameter adjustment complexity, and network robustness. To address these challenges, this paper proposes a clustering optimization algorithm CHNN-CA based on a continuous Hopfield neural network (CHNN). The goal is to enhance the stability and efficiency of FANETs through multi-objective dynamic balance optimization. A CHNN optimization model that incorporates node mobility prediction and energy sensing is constructed to cope with the network’s time-varying characteristics. Additionally, a dynamic penalty function is designed to strengthen cluster size and connectivity constraints, ensuring the convergence of the algorithm under topology uncertainty. Simulation results demonstrate that CHNN-CA achieves a 23.6% improvement in cluster stability and an 18.9% reduction in control overhead compared to the traditional weighted clustering algorithm (WCA) under highly dynamic scenarios. The algorithm also simplifies parameter tuning through a lightweight adaptive function and combines link quality and multi-hop coverage mechanisms to effectively address the problem of cluster head over-density in large-scale networks. This provides an innovative solution for efficient management of dynamic FANETs.
- 6 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72 (1) , 353-360
Continuous Hopfield Neural Network (CHNN)-Based Clustering Optimization Scheme for FANETs
With the rapid development of UAV technology, Flying Ad hoc Network (FANET) has become a core network architecture in military, disaster relief, and other fields due to its potential for efficient communication and collaboration in dynamic environments. However, existing clustering algorithms exhibit significant drawbacks in dynamic adaptability, parameter adjustment complexity, and network robustness. To address these challenges, this paper proposes a clustering optimization algorithm CHNN-CA based on a continuous Hopfield neural network (CHNN). The goal is to enhance the stability and efficiency of FANETs through multi-objective dynamic balance optimization. A CHNN optimization model that incorporates node mobility prediction and energy sensing is constructed to cope with the network’s time-varying characteristics. Additionally, a dynamic penalty function is designed to strengthen cluster size and connectivity constraints, ensuring the convergence of the algorithm under topology uncertainty. Simulation results demonstrate that CHNN-CA achieves a 23.6% improvement in cluster stability and an 18.9% reduction in control overhead compared to the traditional weighted clustering algorithm (WCA) under highly dynamic scenarios. The algorithm also simplifies parameter tuning through a lightweight adaptive function and combines link quality and multi-hop coverage mechanisms to effectively address the problem of cluster head over-density in large-scale networks. This provides an innovative solution for efficient management of dynamic FANETs.