An efficient resource management strategy can save network resources and improve network quality of service (QoS). In order to solve the problem of resource management difficulties of multiple SPs sharing the same backhaul network and QoS mismatch in heterogeneous networks, a downstream resource management in 5G virtualized backhaul network is proposed in this paper. Firstly, the wavelength assignment problem is described as a mathematics model with the consideration of minimizing turning overhead and load balancing. The quantum-behaved particle swarm optimization algorithm is used to solve the mathematics model. Secondly, the 5G traffic is mapping to different priority queues in TWDM-PON by the proposed QoS mapping algorithm. Finally, the downstream traffic is scheduled by the proposed resource scheduling algorithm. Numerical results show that the proposed mechanism can maintain high resource utilization, reduce wavelength tuning overhead and improve the delay of traffic.

In this paper, we propose a model predictive control (MPC)-based approach for the speed tracking control of subway trains, specifically designed to handle the complexities of constrained optimization in real-world applications. The control strategy accounts for nonlinearities, actuator constraints, and disturbances and is validated using real operational data from the Ningbo Metro Line 7. Unlike traditional PID controllers, which are limited by their inability to handle dynamic changes and nonlinear systems, the MPC method optimizes control inputs by predicting future system behavior and adapting to variations in speed profiles. The effectiveness of the proposed MPC-based controller is demonstrated through a comparison with the PID control system already deployed on the metro line. Our results show that MPC provides superior speed tracking performance, particularly in dynamic and disturbed operating conditions, highlighting its potential advantages for real-world subway train control systems.

This paper is concerned with the problem of secure fusion estimation for cyber-physical systems under false data injection (FDI) attacks and eavesdropping attacks. In this work, the active and passive defense mechanisms are respectively designed to preserve the privacy of local estimation information from eavesdroppers and to obtain satisfactory fusion estimation performance against FDI attacks. Specificly, encryption and decryption schemes are designed for the transmitted local estimates, which prevents the eavesdropper from obtaining the correct estimates while the monitoring center is not affected. Then, a secure fusion method is proposed consisting of encryption-based attack detection and prediction based compensation fusion, which can effectively reduce the impact of FDI attack signals. Finally, an illustrative example is employed to show the effectiveness and advantages of the proposed methods.

In wind turbine systems, the state of the generator is always disturbed by various unknown perturbances, which leads to system instability and inaccurate state estimation. In this paper, an intermediate-variable-based distributed fusion estimation method is proposed for the state estimation problem in wind turbine systems. By constructing an augmented state error system and using the idea of bounded recursive optimization, the local estimators and distributed fusion criterion are designed, which can be used to estimate the disturbance signals and system states. Then, the local estimator gains and the distributed weighting fusion matrices are obtained by solving the established convex optimization problems. Furthermore, a compensation strategy is designed by using the estimated disturbance signals, which can potentially reduce the influence of the disturbance signals on the system state. Finally, a numerical simulation is provided to show that the proposed method can effectively improve the accuracy of the estimation of the wind turbine state and disturbance, and the superiority of the proposed method is illustrated as a comparison to the Kalman fusion method.

Real-time, fine-grained gait phase recognition is essential for human– machine synergy in wearable systems. However, this goal faces a persistent accuracy-efficiency trade-off, as capturing complex spatio-temporal-frequency (STF) features from multi-modal biomechanical signals is computationally costly. Traditional single-path methods suffer from representation bottlenecks, leading to coarse-grained recognition. While advanced STF models improve this, their static fusion impairs the recognition of short-duration sub-phases, and their bloated architectures fail to meet stringent real-time inference requirements on edge devices. Therefore, we propose a novel dual-module architecture, called Lightweight Spatio-Temporal-Frequency Fusion Network (L-STFF Net). The network employs a lightweight Spatio-Temporal-Frequency Parallel Feature Extraction Module (STF-PFEM) that utilizes efficient depthwise separable convolutions for concurrent, low-cost STF feature extraction. Subsequently, a Hierarchical Linear Cross-Attention Fusion Module (HLCA-FM), with linear computational complexity, adaptively fuses these multi-domain features, overcoming static fusion limitations. For validation, we collected a dataset comprising surface electromyography (sEMG) and inertial measurement unit (IMU) signals from 12 healthy participants during variable-speed walking. To generate expert-level ground-truth labels cost-effectively, we developed a Semi-automatic Sub-phase Labeling Pipeline (SSLP) that uses plantar pressure and video cues. The L-STFF Net achieved state-of-the-art average accuracies of 97.93% (intra-subject) and 92.16% (inter-subject). Notably, it also attained precise recognition (93.91% recall) on the shortest-duration sub-phase. Furthermore, the network is exceptionally lightweight (66.56k parameters), achieving a single-inference latency of only 9.27 ms on a Raspberry Pi 5. Consequently, L-STFF Net establishes a robust perceptual foundation for next-generation wearable systems by effectively balancing high-fidelity, fine-grained recognition with real-time computational efficiency.