Maximizing the bottleneck bandwidth is a critical challenge for high-capacity data transmission. Existing distributed methods often fail in random topologies because local edge statistics do not necessarily reflect end-to-end path bottlenecks. This paper proposes a distributed routing method based on Ant Colony Optimization with a novel node-based learning mechanism that resolves the limitations of relying on local edge statistics by dynamically aligning the evaluation reference with actual path capacities using a finite history. Simulation results demonstrate a significant performance improvement, with the optimal path selection ratio from approximately 50% to 80% in random networks, while maintaining robustness in dynamic environments like Information-Centric Networking.

In a dense network where numerous stations (STAs) are connected to a wireless network within a designated area, simultaneous data transmissions by multiple STAs frequently cause packet collisions, leading to reduced communication efficiency and increased transmission delays. In wireless local area network (LAN), which is utilized as a wireless network, carrier sense multiple access with collision avoidance (CSMA/CA) is used to avoid packet collisions. In CSMA/CA, a STA observes whether the channel is idle before transmission; if the channel is idle, it waits for a random duration based on the contention window (CW) before transmitting data. However, if multiple STAs select the same waiting time, they transmit simultaneously, resulting in packet collisions. To reduce the probability of packet collisions, it is effective to set a CW that is appropriate for the environment. Nevertheless, in dense networks, packet collisions cannot be avoided even when the CW is set to its maximum value. Therefore, this paper explores the application of p-persistent CSMA, which introduces a transmission probability, and investigates the optimal control of both CW and transmission probability for the environment. The effectiveness of this approach is demonstrated through computer simulations.

Information-Centric Networking (ICN) [1] focuses on discovering and delivering content based on content identifiers. Previous studies utilizing the Self-Organizing Map (SOM) [5] have optimized cache placement and search efficiency, but, deterministic node selection has limited adaptability and flexibility. To address this issue, this study proposes a hybrid search algorithm that combines SOM with Ant Colony Optimization (ACO) [6] to dynamically and efficiently learn optimal routes. In addition, a shared pheromone management scheme based on content attributes is introduced to enhance information sharing among similar content, while reducing memory usage and improving search performance.

To achieve ultra-low latency and power efficiency, metro networks should support a wavelength-direct network by deploying all-photonics technologies. However, conventional reconfigurable optical add-drop multiplexers utilize wavelength selective switches (WSSs) and optical amplifiers to implement colorless, directionless, and contentionless functionality, increasing power consumption and cost. To address this, we previously proposed a two-node cooperative node architecture using low-power contention-based programmable WSSs with passive optical components. In this paper, we investigate the effectiveness and feasibility of the proposed architecture in terms of power-saving, blocking probability, and transmission characteristics. Simulation results demonstrate a 74.7% power reduction, equivalent blocking performance beyond 76 wavelengths, and an optical signal-to-noise ratio of 27.22 dB. We also conducted transmission experiments in a laboratory environment to verify the basic operation of the proposed architecture. We evaluated crosstalk and insertion loss and confirmed that the transmission performance was sufficient.

When a mobile terminal connected with MIMO moves in high speed, the channel capacity is degraded because each element of the channel matrix varies and deviates from the previous reference symbol (RS) and inter-stream or inter-user interference increases in multiuser MIMO. To overcome this problem, Iwakuni, et al. proposed null-space expansion (NSE) method, which forms null space, not only toward the interference channel at the newest RS but also toward those at the past RSs. In this letter, conditional probability density function (PDF) of the varied channel from the one obtained at the previous RS is derived theoretically assuming that each element of the channel matrix varies following i.i.d. (independently and identically distributed) Gaussian process. It is also shown that the expected value of the interfering channel obtained with the PDF is contained in the null space of the array weight by NSE.

This paper proposes a resampler with small computational complexity for differential orthogonal frequency division multiplexing (OFDM) of underwater acoustic communications (UWACs). Although an ambiguity function (AF) method is widely used for resampling in UWACs, it cannot always estimate an appropriate resampling rate from the correlation power if synchronization words (SWs) are Zadoff-Chu (ZC) sequences. The proposed resampler excludes erroneous resampling rates thanks to a detected interval of the two successive SWs. This paper compares the performance of the resamplers based on the AF method, the proposed method and parallel blind resampler selection. Channel simulation results and shallow-water experimental ones demonstrate that the proposed resampler has excellent performance keeping small computational complexity.

This paper presents a comparative evaluation of angle-of-arrival (AoA)-based localization methods using both measured data and simulation results. Several conventional and robust estimators, including least squares (LS), weighted least squares (WLS), weighted linearized least squares (WLLS), two-step error-variance least squares (TELS), and Outlier Sparsity-Promoting Linear Regression (OSPLR), are investigated. While conventional methods provide reasonable performance, they are highly sensitive to outliers caused by measurement errors under severe multipath conditions. Experimental results show that conventional methods achieve comparable accuracy under nominal conditions but suffer severe degradation in the presence of strong AoA outliers. To further interpret these observations, measurement-informed Monte Carlo simulations are conducted by incorporating realistic error statistics and controlled outlier patterns. The results confirm that increasing the number of sensors consistently improves localization accuracy and approaches the Cramér-Rao lower bound (CRLB). In contrast to existing approaches, OSPLR demonstrates superior robustness against severe outliers while maintaining competitive performance under outlier-free conditions. These findings highlight the effectiveness of sparsity-based outlier modeling for practical AoA-based indoor localization in multipath-rich indoor environments.

This letter focuses on a target position estimation using a radar network, which is one of the technologies for recognizing the surrounding environment of a vehicle. The radar network can estimate the target positions based on the ranges from multiple radars to each of targets and existence probability distributions of targets. The position estimation based on the existence probability distributions can be used in multiple-target situations. However, there is a problem of false estimation (ghost targets) at positions where targets do not exist. In this letter, we propose a novel position estimation algorithm employing the simple direction estimation to the targets, which suppresses the appearance of ghost targets. By intentionally offsetting the transmission timing of the two signals from two transmitters, the rough direction of the target is estimated. Then, the estimated direction is used to limit the range of the existence probability distribution. The effectiveness of the proposed method is evaluated by computer simulation.

In this paper, we demonstrate a millimeter-wave active phased array antenna. Our phased array antenna has three features. 1) Planar and back-to-back structure; all electronic components are mounted on the reverse side of the antenna. Our waveguide to microstrip line transition enables the interconnection between both sides of the substrate. 2) Optimized RF-IC placement; the ICs can be placed just behind the transitions in order to reduce attenuation. 3) Low cost; we utilize our original ICs and manufacturing technologies for automotive radars. We designed and fabricated the prototype active phased array with 6 transmitting channels. The radiation patterns were successfully formed both for the boresight beam and for beam scanning. Narrow beam (approximately 2 deg) was obtained. We also confirmed that the waveguide to microstrip line transition did not affect the radiation pattern while the transitions were co-located on the antenna surface.

In monocular depth estimation networks, depth information must be inferred by modeling spatial relationships between points within the image. However, existing networks often struggle to effectively capture both local details and global spatial context. Moreover, due to biases in the dataset, the network learns these biases during training, resulting in limited model generalizability. Therefore, we propose a dynamic hybrid CNN and Transformer network for self-supervised monocular depth estimation. First, we propose a Dual-Dynamic Tanh (DDT) module, a parallel-structured hybrid architecture combining CNN and Transformer. It incorporates a Dynamic Tanh function to enhance self-attention computation, significantly improving the model’s precision for small object prediction. Second, we conducted extensive experiments on both the KITTI and Make3D datasets. Finally, on the KITTI dataset, our proposed network outperforms GeoNet, DF-Net, Monodepth2, LiteMono, and GC-Depth by margins of 8.42%, 7.07%, 2.13%, 0.47%, and 0.70% respectively in the δ<1.25 metric.

Ad hoc multi-hop communication enables flexible connectivity but introduces challenges in securely joining the network without direct access to infrastructure. In such environments, nodes may frequently lose connectivity to an access point (AP) due to mobility or environmental factors, requiring efficient mechanisms to re-establish secure participation without overloading intermediate nodes. Prior work addressed proxy-assisted secure bootstrap using proxy nodes (PNs) and client puzzles to throttle request traffic. However, computational throttling is less suitable in modern environments due to increased processing capabilities and the resulting need for larger puzzle parameters to achieve comparable delays, leading to higher computational overhead and limited adaptability. This letter proposes an authorization-based relay admission scheme, in which an AP issues short-lived relay tokens to previously authenticated nodes to control PN-assisted forwarding during re-bootstrap attempts. A PN forwards a request only when a valid token is presented, while authentication and network admission remain AP-terminated. The proposed approach replaces computational throttling with authorization-based control, reducing unnecessary request propagation while preserving the original bootstrap architecture. Furthermore, the proposed scheme enables lightweight and policy-based control of relay usage without imposing additional computational burden on legitimate nodes.