In recent years, a wide variety of devices, not just smartphones and connected vehicles, have been equipped with communication capabilities. As a result, Internet traffic has increased, and the resulting increase in network peak load has become a problem. Since the increase in peak load causes problems such as temporary decrease in communication throughput, it is considered effective to implement a mechanism to autonomously suppress communication on the terminal side when network congestion is expected. In this paper, we propose a congestion avoidance communication method that suspends data transmission during network congestion and resumes data transmission when the congestion is resolved, using Explicit Congestion Notification (ECN), an explicit congestion notification scheme at the IP layer, for non-time critical applications. We implement the proposed method and evaluate it in an emulation environment. And we also evaluate it in a simulation environment that combines a traffic flow simulator and a network simulator for application to connected vehicles, demonstrating that the proposed method works effectively.

Mobile edge computing (MEC) deploys storage and computing resources at the network edge to handle critical applications that are latency-sensitive and computing-intensive. However, wireless devices (WDs) often face challenges in making efficient offloading decisions in dynamic environments. Additionally, the heterogeneous tasks generated by WDs exhibit varying levels of latency sensitivity, making the traditional approach of optimizing solely for latency inadequate. In this paper, we propose a deep reinforcement learning (DRL) algorithm based on the Actor-Critic (AC) framework to maximize the processing utility of tasks. We consider a time division multiple access (TDMA) system, where data is transmitted using packet communication. The task offloading and resource allocation problem is modeled as a constrained Markov decision process with a hybrid action space. Simulation results demonstrate that the proposed DRL-based utility optimization algorithm achieves faster inference speed compared to deep Q network and effectively improves the task processing rate, achieving a 5% increase in task utility.

This study presents a novel technique to improve axial ratio (AR) of wide-angle scanning phased array. This technique can be applied to phased arrays of reconfigurable dual-polarized circular patch antennas: each antenna element has two feeding channels, with separate variable gain amplifier and phase shifter for each channel. In the proposed technique, compensation weights, excitation amplitudes and phases of each feeding channels, for the AR compensation are controlled based on a closed-form design formula. The design formula is obtained by approximately formulating an active element pattern of the circular patch antenna array. Calculation conditions for the design formula and the effectiveness of the proposed technique have been evaluated by the simulation of a 41-element array. In the simulation, the worst AR in the designated coverage range of ±70^° was improved from 5.8 dB to 2.0 dB by the proposed technique. We experimentally tested the proposed technique using a 1024-element phased array. The measured AR was improved from 5.8 dB to 1.9 dB at 70^° scan by the proposed technique.

The rapid deployment of satellite constellations has propelled the swift development of the satellite industry. Low-cost solutions spanning the entire industry chain will significantly accelerate the improvement of satellite internet. On the user side, phased array antenna terminals are gradually replacing reflector antennas, yet the ensuing challenges of hardware architecture and costs. Lowering the resolution of phase shifter is one solution, but this may result in a loss of beam pointing accuracy. A novel two-stage phase shifter phased array architecture is proposed, coupling subarray division with mixed-resolution phase shifters to reduce costs while ensuring the beam performance. The architecture boasts low complexity and ease of implementation. Simulation results demonstrate the balanced performance of this architecture in scanning accuracy and beam sidelobes, marking a promising path for industrial applications.

This study introduces a three-dimensional (3D) complex permittivity profile reconstruction using a deep neural network, where wave-number space data compression is applied to reduce the dimension of input data. Four-dimensional scattered data are converted into a 3D complex permittivity profile by integrating a 3D convolutional autoencoder and a multilayer perceptron. The reconstruction accuracy is further improved through efficient skin surface rejection preprocessing via a fractional derivative model. An experimental study, using simplified 3D breast phantom and an ultrawideband radar module shows that our proposed scheme provides accurate estimates for 3D reconstruction in terms of relative permittivity and conductivity.