High torque ripple dramatically affects the switched reluctance motor (SRM) application. To reduce the torque ripple, a reference torque neural network (RTNN) is proposed to adjust the reference torque online. Firstly, the RTNN is built on the torque sharing function (TSF) method. Furthermore, the RTNN is designed as a single-input and -output network. As the periodic relationship between the torque ripple and the rotor angle, the rotor angle constitutes the central node parameter of the implicit function in RTNN. Therefore, one-step adjustment of the RTNN can perform well at restraining reference torque. Lastly, the torque error is used to adjust the parameters of RTNN to reduce the torque ripple. In the MATLAB environment, through the simulation comparison with fuzzy torque and PD current compensation method, the effectiveness of RTNN at torque ripple suppression is proven with different loads and speeds.

Dynamic complex matrix inversion (DCMI) problems frequently arise in the territories of mathematics and engineering, and various recurrent neural network (RNN) models have been reported to effectively find the solutions of the DCMI problems. However, most of the reported works concentrated on solving DCMI problems in ideal no noise environment, and the inevitable noises in reality are not considered. To enhance the robustness of the existing models, an anti-noise parameter-variable zeroing neural network (ANPVZNN) is proposed by introducing a novel activation function (NAF). Both of mathematical analysis and numerical simulation results demonstrate that the proposed ANPVZNN model possesses fixed-time convergence and robustness for solving DCMI problems. Besides, a successful ANPVZNN-based manipulator trajectory tracking example further verifies its robustness and effectiveness in practical applications.

Batched sparse (BATS) coding is a promising new technique designed for reliable transmissions over erasure networks, which can achieve asymptotic bandwidth optimality. However, the decoder of the conventional BATS codes begins to recover the source packets only after receiving a sufficient number of encoded packets which induces delay. Moreover, the larger the source block size is, the longer the time delay will be. In this paper, we propose a scheme by using expanding window applied over BATS codes, termed the expanding-window BATS codes, which divide the source data into smaller blocks for reducing delay. Furthermore, for improving the decoding performance we design an adaptive encoder which encodes the source packets with dynamic encoding probabilities. Because the performance of BATS codes highly depends on the degree distribution, we further optimize the degree distributions for the proposed expanding-window BATS codes with linear programming and the performances have been verified by the simulation results.

Cognitive radio combined with non-orthogonal multiple access (CR-NOMA) networks is a research hotspot in the wireless communication field. All the time, secure communication of wireless networks has become a pressing issue due to the openness of the wireless channel. Secondary outage is one of the important secrecy performance for the wireless communication. Secrecy outage performance (SOP) for a CR-NOMA network with three secondary users (SUs) was firstly investigated in our last work, which is still very rare so far. However, an assumption is only Rayleigh fading for the last work, which is not a practical model for the wireless communication networks. In the paper, we firstly introduce a CR-NOMA networks with three SUs over Nakagami-m channels, which utilizes multi-antennas technology at the base station (BS) to enhance SOP performance. In particularly, we propose two appropriate transmitter antenna selection schemes which are respectively a maximum channel capacity (MCC) scheme and a maximum-minimum (MM) scheme to optimize resources and mitigate common eavesdropping scenarios. The MCC scheme enhances system SOP performance under passive eavesdropping, while the MM scheme improves system SOP performance under active eavesdropping. Subsequently, we derive precise closed-form SOP expressions under the above two schemes. In order to obtain deeper insights, we analyze their asymptotic SOP performance. The asymptotic results indicate that the power allocation factor significantly influences system SOP performance. We utilize a simulated annealing algorithm to find the optimal power allocation factor which further enhance the system SOP performance. The results reveal that the optimal power allocation factor is not fixed, which changes with the transmit power of the BS. Finally, we verify the theoretical analysis through Monte Carlo simulations. Simulation results demonstrate that the MM and MCC schemes achieve higher SOP performance (6%-55% when transmit power is 20 dBm) and secrecy throughput than the single antenna. Meanwhile, the system SOP performance under the Nakagami-m channel is better than that under the Rayleigh channel.

The reconfigurable intelligent surface (RIS) has garnered significant attention in recent years due to its cost-effectiveness and ability to reconfigure the wireless environment in a controlled manner. However, signals reflected by the RIS need to traverse two paths—from the base station (BS) to the RIS and from the RIS to the user, which results in the multiplicative fading effect and limits the capacity gains achievable by the passive RIS. In this paper, we investigate the active RIS to overcome the limitation. We consider a multiple input single output system utilizing active RIS, and based on recent advancements in deep reinforcement learning (DRL), we integrate the joint design of the BS transmit beamforming matrix and the RIS phase shift matrix into a unified framework for sum rate maximization. We propose a DRL-based algorithm that obtains the joint design through trial-and-error interactions with the environment and is capable of identifying an appropriate power allocation scheme to balance the power consumption between the BS and the active RIS. Simulation results demonstrate that the proposed active RIS-assisted system outperforms traditional passive RIS-assisted systems, particularly when there are constraints on the number of reflective elements. Additionally, it is observed that proper parameter settings significantly enhance the performance of the proposed algorithm. Finally, the algorithm can allocate suitable power to the active RIS while maintaining a constant total power, thereby optimizing the system performance.