All-Photonics Network (APN) promises to offer high-capacity and low-latency communications by providing direct optical paths between terminals. Furthermore, APN based on the direct optical paths aims to achieve networking with lower power consumption by minimizing electrical processing. An effective way to increase the provided number of paths and expand the provided area is to introduce an any-to-any wavelength conversion function into an optical node. The key device implementing the function into an optical node is a tunable wavelength converter (WC) with low power consumption. An optical-analog-optical (OAO) configuration consisting of a cascaded coherent receiver and I/Q modulator is promising for low-power tunable wavelength conversion because its configuration omits digital signal processing (DSP) devices. In this paper, we propose applying OAO-WCs to optical nodes and report on experiments focusing on the WCs themselves and the optical nodes incorporating them. Since the OAO-WC does not perform DSP on the signal, we experimentally describe the impact of wavelength conversion on the optical signal characteristics using OAO-WC prototypes. Moreover, toward applying the OAO-WCs to optical nodes, we show that the OAO-WCs allow for 2000-km-class wavelength division multiplexing (WDM) transmission with 6 wavelength conversions. This result demonstrates the proposed nodes are applicable to core networks, where WDM signals are transmitted on the order of 1000 km via multiple optical nodes. In addition, we demonstrate the physical feasibility of optical nodes consisting of optical transmission systems and the OAO-WCs by showing that the additional Q-penalty was negligible when integrating them. This result indicates that already installed optical nodes can be upgraded to have an any-to-any wavelength conversion function simply by adding the OAO-WCs, which can contribute to smoothly introducing the proposed node into existing networks.

In high-symbol-rate and ultra-dense wavelength-division-multiplexing (WDM) systems, spectrum narrowing poses a critical impairment due to the limited bandwidth of digital-to-analog converters (DACs), analog-to-digital converters (ADCs), and the non-ideal passband characteristics of wavelength selective switches (WSSs). Although the resulting inter-symbol interference (ISI) can be compensated by adaptive equalizers, such compensation inevitably enhances noise and destabilizes digital signal processing (DSP). We propose a demodulation framework that integrates opto-electrical hybrid partial-response spectrum shaping with recurrent neural network (RNN)-based sequence estimation to mitigate spectrum narrowing. By treating transmission-induced spectrum narrowing as part of a partial-response filter (PRF), the proposed scheme suppresses noise enhancement, stabilizes adaptive equalization, and enables sequence estimation without relying on channel information. Comprehensive numerical simulations, covering various modulation formats and transmission distances, confirm the effectiveness of the proposed scheme. Its feasibility is further demonstrated through transmission experiments using 32-Gbaud dual-polarization 4-QAM signals spaced at 37.5 GHz, corresponding to ultra-dense WDM scenarios.

A novel polarization-maintaining fiber (PMF) with fluorine-doped stress rods is proposed. Unlike the conventional PMF with boron doped stress rods to produce birefringence from thermal stress, the new design relies on birefringence induced by mechanical stress, which is temperature insensitive. A fiber is designed and fabricated. It exhibits optical properties compatible with standard single-mode fiber. Fiber beat lengths of 26.7 mm at 1310 nm and 32.8 mm at 1550 nm are achieved. Fiber polarization extinction ratios of 41 dB/m at 1310 nm, and 29 dB/m at 1550 nm are demonstrated. A mechanical-optical model is developed to analyze the stress distribution and quantify the fiber birefringence. The modeled fiber beat lengths agree with the measured beat lengths of the experimental fiber.

To satisfy the ultra-low latency requirements of Beyond 5G, realizing direct user-to-user wavelength connections in metro networks is crucial. We previously addressed this by proposing a multiband massive wavelength-multiplexing metro network (M3 Wavenet), an architecture leveraging energy-efficient contention-based wavelength selective switches (WSSs) to accommodate massive multiple wavelengths. However, its effectiveness has not yet been fully verified under diverse scenarios. Therefore, in this paper, we investigate the impact of remotely located transceivers in M3 Wavenet on network performance, which is essential for user direct wavelength connection. Through extensive simulation experiments, we quantify the resource efficiency in terms of connection blocking probability and the techno-economic viability regarding network cost. The results demonstrate that the remotely located transceiver model can achieve acceptably low blocking probability (less than 10^-2) with 128 wavelengths per fiber. Furthermore, deploying contention-based WSSs instead of contentionless WSSs and limited-range tunable lasers can reduce network cost, but there is a concern that the lack of flexibility may degrade resource efficiency. So, we also investigate the performance and cost benefit of such devices. We demonstrated that the proposed architecture reduced the network cost by about 70% in comparison with the conventional architecture. We confirmed the techno-economic viability of the remotely located transceiver model in terms of both blocking probability and network cost in various network conditions.

To address the rapid increase in communication traffic, optical backbone networks are focusing on technologies such as Elastic Optical Networks (EONs) and Space Division Multiplexing (SDM), which enable higher-speed and higher-capacity transmission. In SDM-enabled EONs employing multi core fibers (MCF), inter-core crosstalk (XT) becomes a performance limiting factor. Although the previous study has proposed spectrum allocation methods for distributed-control SDM-EONs employing 4-core MCF, the impact of XT has not been sufficiently addressed. This paper proposes an XT-aware resource allocation method for distributed-control SDM-EONs employing a 7-core MCF under fixed shortest-path routing. The proposed method consists of two mechanisms. First, it employs an XT-aware slot allocation algorithm that switches between First-Fit and Last-Fit according to the XT tolerance of each optical path. Second, it adopts a modulation format selection algorithm that considers XT tolerance and spectrum requirements. By combining these mechanisms, the proposed method mitigates XT-induced allocation constraints and improves spectrum utilization. Simulation results show that the proposed method improves the traffic capacity until the call blocking probability satisfies 10^-3 by up to 14.2% compared with the conventional method.

This paper introduces a novel approach to semantic communication (SC) that utilizes generative Artificial Intelligence (AI) models, namely, stable diffusion, to transmit images with significantly reduced bandwidth consumption while maintaining high reconstruction fidelity. While our previous work explored the potential of multi-dimensional semantics, this paper presents a robust framework specifically designed for the demands of vehicular networks. We introduce a comprehensive set of 13 complementary semantic features that capture fine-grained scene details, enabling faithful image reconstruction even in bandwidth-constrained scenarios. This structured representation significantly enhances reconstruction fidelity while maintaining the system’s adaptability to diverse receiver preferences. We show the effectiveness of this extended framework, which achieves high reconstruction fidelity with a Fréchet Inception Distance (FID) of 17.24 and reduces the transmitted data size by over 92% compared to the original image, while retaining the key capability of customized, style-varied reconstructions for different receivers.

Adopting artificial intelligence (AI) and machine learning (ML) based algorithms as fundamental building blocks of the next generation radio access networks (RAN) has been a subject of intense study in the recent years. In particular, more extensive use of AI/ML solutions is envisioned as one of the key aspects of the future 6G radio systems. In this work, we focus on one promising use case for AI in the air interface, where an ML receiver is trained to detect transmissions without channel estimation pilots, instead relying on a simultaneously learned constellation shape. Based on extensive over-the-air measurement campaigns, we quantify the performance gains from such an approach in different environments. In addition, we also address the commercial feasibility of the proposed approach and AI-RAN concepts in general, reflecting on the achieved performance results.

In 6th-generation mobile communication system (6G), intelligent network optimization, resource allocation, and spectral efficiency can be improved by implementing artificial intelligence (AI)/machine learning (ML) in communications, networks, and devices. A concept called AI-native Air Interface (AI-AI), which applies AI/ML technology to the air interface of mobile communication, has been proposed for the realization of 6G intelligent networks. This concept aims to provide data transmission required for mobile applications in the most efficient way by optimizing the air interface end-to-end, considering both hardware capability and radio propagation environments. A pilotless transmission scheme based on deep learning has been proposed as one of the technologies to realize AI-AI. We have demonstrated the effectiveness of this technology in indoor fixed-point and low-speed mobile environments using experimental equipment. However, the pilotless transmission scheme should be evaluated in a variety of fading channel models. In this paper, we conduct experimental trials in fading environments generated by a fading channel emulator and evaluate the transmission performances of the pilotless transmission scheme. The experimental results show that the scheme is more effective in high-speed mobile environments than in low-speed mobile environments.

The use of power domain non-orthogonal multiple access (NOMA) in cell-free massive multiple-input multiple-output (CF-mMIMO) may further improve the spectral efficiency (SE). In this paper, we consider a multiple central processing unit (CPU) environment to decentralize the CPU processing. However, when each CPU executes user clustering (UC) independently, the user terminal (UT) experiences interference from the access points (APs) controlled by the neighboring CPUs. A coordinated multi-CPU scheme, in which each CPU controls the APs of the neighboring CPUs that may interfere with UTs, can reduce the interference, but the computational complexity of the UC algorithm increases as the number of controlled APs increases. In this paper, we analyze the trade-off between the effectiveness of interference reduction and the increasing overhead owing to coordinated multi-CPU by introducing the effective resource ratio, which represents the ratio of the time-frequency resources used for data transmission within the resources managed by CPUs. The analysis results reveal that lower CPU density increases overhead and reduces the SE of UTs under coordinated multi-CPU environments. Conversely, higher CPU density improves interference reduction effects and increases the SE of UTs.

This paper proposes a robust inter-user interference (IUI) suppression scheme for multi-user massive multiple-input multiple-output (MU-mMIMO) with multi-stream signaling in time-varying channel environment. The MU-mMIMO system enhances spectral efficiency by multiplexing several streams in spatial domain. In the mobility environment of user terminals (UTs), MU-mMIMO transmissions are subject to both IUI and inter-stream interference (ISI). Null-space expansion (NSE) exploits degrees of freedom (DoFs) of base station (BS) and enhances the IUI suppression capability by steering several nulls to interfering UTs. Conventional techniques based on NSE assume that each terminal operates on a single stream. Focusing on multi-stream transmission per UT with multiple antenna elements, this paper newly designs NSE based on block diagonalization (BD). We investigate the application of NSE that balances high interference suppression performance with low computation complexity, alongside various BD algorithms. Their effectiveness are demonstrated through computer simulations, ultimately identifying the optimal BD-NSE configuration for mobility environments.

We are studying NRNT (New Radio Network Topology) which enhances MIMO (multiple-input and multiple-output) spatial multiplexing and communication capacity in 6G system using spatially distributed devices such as RIS (Reconfigurable Intelligent Surface) which can control the direction of reflected waves electrically. On the other hand, hybrid BF (beamforming) is necessary to compensate the large propagation losses at high frequencies. It is necessary to optimize the beam selection of each antenna to achieve MIMO spatial multiplexing, but the number of beam combinations exponentially increases when using multiple RISs. In this paper, we propose a fast beam-selection algorithm for a hybrid beamforming system and discuss its computational complexity. We also conduct simulations based on the 3GPP propagation model and show that using the proposed algorithm increases throughput.

Sixth-generation mobile communication system (6G) aims at extremely high-data rate and large-capacity communication over 100 Gbps which surpasses fifth-generation mobile communication system (5G). At the same time, the use of 7 GHz to 24 GHz frequency band called centimeter wave (cmWave) or FR3 is also considered to ensure the coverage for the 6G systems. In addition, hybrid beamforming (BF) in Massive MIMO systems is considered to compensate for the propagation loss considering the implementation cost and inter-stream interference. In this paper, by using computer simulations, the transmission characteristics and coverage performance of cmWave Massive MIMO systems are evaluated. In the evaluations, the beam control, which can make the best use of hybrid BF, were evaluated using 3GPP channel model and urban model. From the simulation results, it was clarified that the transmission characteristics of Sub-6 or higher frequency bands and that cmWave could provide more stable communication than millimeter wave. In addition, when applying hybrid BF to cmWave Massive MIMO, it was found that it could reduce the computational complexity compared with high frequency bands by utilizing the wide beam width.

This paper presents a reliability-ensuring platform designed to support stable cybernetic avatar (CA) teleoperation, as part of Japan’s Moonshot Goal 1 project. We develop a Multiple QoS (MQoS) node with three core functions to ensure end-to-end (E2E) communication reliability: (1) monitoring and visualizing of communication quality, (2) jitter buffering, and (3) dynamic path switching. To evaluate the effectiveness of the proposed platform, we conducted experiments in real network environments. In a teleoperation setup between two sites approximately 50 km apart, the jitter buffering function successfully reduced jitter (delay variation) to approximately 4 ms under degraded communication conditions. Furthermore, we measured E2E delay and jitter between Yokosuka and multiple domestic and international sites using various network operators. While average delay increased with distance and routing complexity, average jitter remained relatively stable across locations. However, characteristic spike jitter phenomena were observed under specific combinations of locations and network operators. Based on these observations, we also utilized a jitter model proposed in [5] to reproduce such spike jitter behaviors in the simulated environment, and confirmed that similar communication degradation and its impacts on CA teleoperation can be reproduced. These results demonstrated that the proposed MQoS-based platform can effectively monitor, mitigate, and adapt to communication quality variations in heterogeneous real-world networks, contributing to stable CA teleoperation. In future work, we plan to extend the validation to a broader range of CA types and teleoperation scenarios, and to improve scalability and automate MQoS node configuration for practical deployment in dynamic network environments.

Cooperative operations among multiple unmanned underwater vehicles (UUVs) are emerging as a prevailing trend for executing complex marine missions. Accordingly, the multi-UUVs mobile underwater acoustic networks, which serve as the foundation for information exchange between UUVs, have become a key research focus. However, the high mobility of UUVs causes dynamic node positions, leading to unstable channels and time-varying topologies and excessive network maintenance overhead. To address these problems, a Multi-Dimensional Spatio-Temporal Evolution Network graph prediction-based Greedy Perimeter Stateless Routing protocol (MDSTN-GPSR) is proposed to achieve more reliable routing in multi-UUV mobile swarm networks. To address the issues of routing holes and detours caused by the high mobility of UUVs, a multi-dimensional spatio-temporal evolution network graph prediction model is proposed to construct the time-varying network location maps, topology, and conflict graph for routing calculation. A Low-Overhead Neighbor-Aware Route Maintenance Mechanism is introduced to reduce the scope and frequency of routing, thereby minimizing overhead. The proposed method uses a Cross-Layer Optimization-based Multi-Factor Forwarding Mechanism. This mechanism selects paths based on both physical link quality and MAC scheduling slots, in order to improve the delivery rate and reduce end-to-end delay. Simulation results show that, compared with the GPSR routing protocol, MDSTN- GPSR reduces average end-to-end delay by 10.14% and increases packet delivery rate by 20.26%, adapting better to the time-varying topology of multi-UUV underwater mobile swarm networks.

Service Function Chaining (SFC) establishes service paths where intermediate nodes execute network functions in a specified order. Integrating Cloud-Native (CN) technologies enhances SFC’s agility, flexibility, scalability, and fault tolerance by leveraging Cloud-native/Containerized Network Functions (CNFs) and their orchestration. Network Service Mesh (NSM) extends Kubernetes’ networking capabilities and enables effective NSM-based CN-SFC implementations. Several studies have shown NSM-based CN-SFC (NSM-SFC) improves deployment time, end-to-end latency, and failure recovery. However, automatic SFC scaling for dynamic load fluctuations remains underexplored. In this paper, we propose Horizontal SFC Autoscaler (HSA) for NSM-SFC, which dynamically adjusts the number of exclusive service paths based on service demand. HSA integrates Kubernetes’ autoscaling function, Horizontal Pod Autoscaler (HPA), with NSM’s Topologically-aware Scale from Zero (TSZ) mechanism to achieve path-level autoscaling. Unlike the conventional TSZ, which ties CNFs to specific deployment nodes, HSA ensures exclusive service paths by associating CNFs with their respective NSM clients, uniquely identified to maintain path exclusivity. Experiments demonstrate that HSA efficiently scales resources to meet demand and improves fault isolation by creating disjoint service paths. We also reveal that the current TSZ mechanism requires improvement to reduce recovery time.

Toward 6G and further enhancement of 5G, metasurface and reconfigurable intelligent surface (RIS) attract much attention thanks to their cost-effective and energy-efficient features. Some experimental trials have been conducted to verify their effectiveness. In consideration of a legacy deployment where a base station (BS) is deployed at the rooftop of a building, it is difficult to create a coverage area at the foot of the building in high frequency bands such as millimeter-wave (mmW) bands. Thus, for enhancing the coverage and system capacity, we propose novel solutions using multiple transmissive metasurfaces (TMSs) combined with beamforming (BF) in a mmW-band BS deployed in indoor. To verify the effectiveness of the proposed solutions, 5G experimental trials using a 28 GHz-band BS and a user equipment (UE) were conducted. To expand the coverage, the BF function selectively aims the beam at one of five TMSs, and then TMS re-directs the radio wave to the foot area. On the other hand, to enhance the system capacity, the beam is focused on UE located at the foot of the building through five TMSs. This paper introduces an overview of the 5G experimental trials and shows measurement results of received power and downlink throughput in the trials. It is verified that our solutions can enhance the coverage area to the foot and achieve much higher throughput than the measurement results in the legacy deployment and the indoor deployment without any TMSs.

This paper investigates uplink grant-free access (GFA) for massive connectivity in the presence of carrier frequency offsets (CFOs) and proposes a new independent component analysis (ICA)-based receiver that integrates differential phase-shift keying (DPSK). The conventional ICA-based classifier-diversity-combining (CDC-ICA) scheme executes multiple ICA classifiers in parallel, each of which blindly estimates the transmitted signals and performs active user detection (AUD) by decoding the coded identification (ID) sequences. Although this scheme can perform AUD and data decoding while mitigating pilot overhead, perfect frequency synchronization between the base station and all users is assumed, which is impractical in GFA scenarios. To this end, the proposed scheme employs a CFO-robust active user enumeration and DPSK-aided phase rotation correction to mitigate the effect of CFOs. Furthermore, the CDC-ICA scheme takes the union of the results obtained from multiple ICA classifiers and, as a result, tends to accumulate false alarms as the number of classifiers increases, resulting in unnecessary data decoding for inactive users. To address the issue, we further consider two thresholding strategies: one based on the log-likelihood-ratio (LLR) and another based on majority voting. Simulation results demonstrate that, under CFO, the proposed scheme achieves a lower bit error rate (BER) and miss-detection probability (MDP) than the conventional schemes, while effectively balancing the trade-off between false-alarm probability (FAP) and BER.

This paper proposes BIADS, a bio-inspired adaptive dialogue framework for multimodal, state-aware conversational support. BIADS builds a dynamic user profile from text, voice, and visual signals to estimate evolving support-related states and to guide graded intervention. Its core is a phased alert protocol, functionally inspired by plant defense logic, that regulates support intensity according to inferred need. Experiments on the MultiDACS dataset show that BIADS outperforms strong baselines in response selection (R@1=0.783), supportive intervention effectiveness (State Mitigation Rate =0.89), and human-rated empathy. These results suggest that combining multimodal state profiling with staged support control is effective for adaptive dialogue systems.