This paper addresses the challenge of anti-jamming in moving reactive jamming scenarios. The moving reactive jammer initiates high-power tracking jamming upon detecting any transmission activity, and when unable to detect a signal, resorts to indiscriminate jamming. This presents dual imperatives: maintaining hiding to avoid the jammer’s detection and simultaneously evading indiscriminate jamming. Spread spectrum techniques effectively reduce transmitting power to elude detection but fall short in countering indiscriminate jamming. Conversely, changing communication frequencies can help evade indiscriminate jamming but makes the transmission vulnerable to tracking jamming without spread spectrum techniques to remain hidden. Current methodologies struggle with the complexity of simultaneously optimizing these two requirements due to the expansive joint action spaces and the dynamics of moving reactive jammers. To address these challenges, we propose a parallelized deep reinforcement learning (DRL) strategy. The approach includes a parallelized network architecture designed to decompose the action space. A parallel exploration-exploitation selection mechanism replaces the$\varepsilon $-greedy mechanism, accelerating convergence. Simulations demonstrate a nearly 90% increase in normalized throughput.

Data-driven cross-layer secure design is expected to provide effective support for high-reliability and high-speed 6G network services. In this paper, we study the joint anti-jamming decision problem for phy-layer scheduling and congestion control in the transport layer. To address the challenges of the extremely huge action space and the simultaneous existence of multi-dimensional and multi-scale action variables, we propose a hierarchical DRL anti-jamming algorithm. Firstly, we unify the time scales of the variables by defining the state space, action space, and reward function, and construct them as a Markov decision process. Second, we make decisions in different dimensions sequentially through a hierarchical learning algorithm, which compresses the size of the action space while maintaining the correlation between variables. Simulation results show that the proposed cross-layer design method can realize a good adaptation between the lower layer and the transport layer in the context of anti-jamming requirements, which significantly improves the user QoS and system throughput compared with the baseline algorithms.

The cover image is based on the Research Article Joint Channel and Power Optimization for Multi-user Antijamming Communications: A Dual Mode Q-learning Approach by Xiaobo Zhang et al., https://doi.org/10.1049/cmu2.12339

Abstract In view of the anti‐tracking‐jamming problem, traditional online learning methods usually cannot analyse the jamming behaviour, and find an effective way to prevent the jamming attacks. To cope with these challenges, a novel communication/deception dual mode mechanism is proposed in this paper. Deception users are selected to send high‐power signal for jamming attraction, and form collaborative relationships with communication users. The corresponding collaborative anti‐jamming model is then constructed as a Markov game to analyse the multi‐agent decision. Based on that, a joint channel and power optimisation for multi‐user anti‐jamming communications based on dual mode Q‐learning scheme is proposed. Compared with two traditional online learning algorithms, the proposed DCAJ‐QL algorithm effectively achieves 146.5% and 80.4% higher maximum communication rate under tracking jamming conditions, and achieves 40.7% and 53.6% higher maximum communication rate under fixed jamming conditions.

To achieve high-performance sensorless control of interior permanent magnet synchronous motor (IPMSM) drives, this paper proposes an extended flux observer by using the current model to modify the voltage model. The current model is set to correct the voltage model through the PI regulator to form closed-loop control. The proposed closed-loop control method effectively suppresses the DC offsets and integral saturation in the voltage model, which greatly improves the stability and accuracy of sensorless control. On this basis, for the 5th and 7th harmonics caused by the inverter nonlinearity, this paper proposes a filter network composed of multiple second-order generalized integrators (Multiple SOGIs), and uses the integral characteristic of the second-order generalized integrator to replace the pure integral of the voltage model. Multiple SOGIs greatly filter out the high-order harmonics in the back electromotive force (EMF) and improve the accuracy of sensorless control. The effectiveness of the proposed scheme is verified on a 2.2kW IPMSM drive platform.

This paper mainly investigates the unmanned aerial vehicle (UAV) communication networks under the threat of a static malicious jammer. While taking the flying process of the user (UAV transmitter-receiver pair) into consideration, we propose a joint power and trajectory optimization method. Moreover, a Stackelberg framework is formulated to solve the proposed optimization problem. In addition, a joint power and trajectory optimization algorithm (JPTOA) based on best-response (BR) is designed to obtain the user's strategy as well as Stackelberg equilibrium (SE) in each time stage. Furthermore, as an extension of single stage optimization, the I-step exploration and one-step decision (IEOD) scheme is designed to enhance the user's cumulative utility. Finally, simulation results are presented to show the performance of the proposed JPTOA scheme.

To better solve the optimization problem of fuzzy neural network (FNN), a kind of method based on artificial fish swarm algorithm (AFSA) is proposed in this paper. Aiming at the structure optimization problem of FNN, AFSA-FNN1 is established and realizes the simplification of fuzzy rules. Aiming at the parameter optimization problem of FNN, AFSA-FNN2 is built and realizes the acquisition of parameters of membership function (MF) automatically. The proposed method uses for path planning of the robot, simulation results show that the optimized FNN can enhance the smoothness of the path.