We demonstrate a far-UVC optical wireless communication system employing $\mu$ LED transmitters emitting at 235 nm. Reducing the device dimensions to the micro-metre regime decreases the active area, enabling higher current densities and lower junction capacitance, which together enhance the modulation bandwidth relative to conventional large-area AIGaN LEDs. The resulting improvement in high-speed performance allows the intrinsic bandwidth limitations of deep-ultraviolet emitters to be mitigated. Using these $\mu$ LEDs in a short-range free-space link, data rates of up to 1.5 Gbps are achieved. These results highlight the potential of far-UVC $\mu$ LED technology for compact, solarblind optical wireless communication systems requiring high data throughput.
This paper presents a novel architecture for realizing a fully parallel and flexible digital baseband processor for massive MIMO systems using processing-in-memory (PIM). The proposed approach leverages memristor-based crossbar arrays to implement key functional blocks within the baseband processing chain, thereby alleviating the data-movement bottleneck inherent in conventional von Neumann architectures. To enable efficient hardware mapping, the signal processing algorithms and corresponding computations are reformulated into a unified matrixbased framework. The resulting matrix operations are then mapped onto multiple crossbar arrays, exploiting their intrinsic analog computing capabilities. The proposed PIM-based baseband processor achieves fully parallel operation and significantly reduces processing latency, making it well suited to the stringent performance requirements of next-generation wireless systems. The presented architecture demonstrates the potential of PIM technology as an effective solution for future high-performance and energy-efficient baseband processing platforms.
This paper evaluates the performance of different multi-point transmission schemes in a software-defined multi-cell Visible Light Communication (VLC) system. Using a GNU Radio and Universal Software Radio Peripheral (USRP)-based testbed with phosphor-converted white LEDs and black silicon photodetectors, we assess the achievable data rates for Zero-Forcing (ZF) precoding and Joint Transmission (JT) scheme in a distributed Multiple-Input Multiple-Output (MIMO) VLC configuration. OFDM waveforms are utilized and co-channel interference is modeled as additive white Gaussian noise. Experimental results show that the data rate performance is mainly limited by the narrow electrical modulation bandwidth of phosphor-converted LEDs, yielding a $150-160 ~\text{cm}$ single-cell coverage radius. In a two-cell scenario with an LED separation of 170 cm, ZF, JT, and Single-Input Single-Output (SISO) are the most convenient scheme for 46%, 32%, and 22% of the service area, respectively.
Using 3D scanning radars is essential for aircraft. MIMO radars significantly simplify their installation in aircraft while significantly increasing equivalent resolution. However, implementing MIMO elements in wings poses a significant challenge due to their deflection, which degrades the radar images or makes it unusable. The objective of our study was to research and develop a method for configuring 3D imaging MIMO radars in the wings of various aircraft, taking into account arbitrary deflections. We examined radar images obtained by radar due to arbitrary deflections of aeroplane wings. Deflection includes both static and dynamic effects, particularly flutter. A mathematical model of a radar installed in deformable wings has been developed. We propose a method for compensating the MIMO model calculation for deflections, which restores the original radar image. We provide a calculation of the accuracy of the method depending on the accuracy of the sensors. We demonstrate that using simple sensors that measure arbitrary wing deflection, high-quality radio imaging can be obtained using MIMO radar. The proposed method is invariant to both the radar frequency range and the aircraft size.
A realistic indoor office in Oxford is modeled to examine channel propagation based on visible light communications (VLC), employing light-emitting diodes as optical luminaires. A design tool for optical systems, called Zemax, is used to model the indoor office room. Moreover, ray tracing in Zemax is accounted for two comparative VLC propagation scenarios. Further, channel impulse response and channel frequency response metrics are emphasized. Building upon these propagation metrics, ultimately, the 3 dB channel bandwidth versus receiver position results are investigated for an indoor office room, accounting for the considered comparative propagation scenarios. Concerning the key findings of scenario 1, 50 MHz and 150 MHz 3 dB channel bandwidths are obtained at the center and at the corners, respectively. Likewise, for scenario 2, 40 MHz and 50 MHz 3 dB channel bandwidths are obtained at the center and at the corners, respectively.
A wavelet-first preprocessing method is proposed for Optical Wireless Communication (OWC) links employing intensity modulation and direct detection with On-Off Keying (IM/DD-OOK) under compound degradation. Attention is focused on the joint presence of slow baseline drift and intermittent impulsive contamination before threshold-based detection in open optical environments. The drift is estimated through the Stationary Wavelet Transform (SWT), whereas impulsive samples are treated on the residual by means of a segmented Hampel-type detector. An optional translation-invariant denoising stage based on cycle-spinning (CS) is then applied to reduce residual noise while limiting edge attenuation in short OOK-like events. Evaluation is carried out on photometric time series from the All-Sky Automated Survey for SuperNovae (ASAS-SN), used as realistic temporal proxies within a controlled semi-synthetic framework for proof-of-concept validation of receiver-side preprocessing under compound degradation. Event-preservation, drift-suppression, impulse-related, and communication-oriented metrics are reported. Improved drift removal and greater robustness to impulsive contamination are observed, while event contrast is preserved and communication performance remains close to the strongest robust baselines.
Dense urban areas have challenges for sixthgeneration (6G) Low Earth Orbit (LEO) Non-Terrestrial Networks (NTN), including signal blockage, atmospheric attenuation, and beam misalignment. To address these issues, this paper an analysis of a Quantum Deep Reinforcement Learning (QDRL)based Reconfigurable Intelligent Surface (RIS) assisted optical beamforming model for dense urban 6G LEO-NTN systems. The proposed model jointly optimizes RIS phase shifts, optical beam steering angles, and transmit power allocation using a quantumenhanced policy optimization method. A hybrid state-action model is developed considering satellite altitude, azimuth and elevation angles, line-of-sight (LOS), non-line-of-sight(NLOS), and urban atmospheric attenuation. The QDRL agent maximizes a long-term reward function formulated over performance evaluation parameter of Signal-to-Interference-plus-Noise Ratio (SINR). The optical channel modeled using free-space optical (FSO) propagation with atmospheric attenuation, geometric spreading, and RIS-induced phase reconfiguration gains. Simulation results of the QDRL-RIS optical beamforming model best performance as compared conventional DRL and static RIS models of SINR, and coverage probability. The model mitigates urban blockages and environmental impairments while improving speed and quantum policy optimization. These results confirm the proposed model for enabling reliable, high-capacity in future dense urban 6G LEO-NTN systems.
We demonstrate synchronous real-time optical performance monitoring of microcomb-based WDM signals via a Fourier-domain optical vector oscilloscope. Frequencyto-time mapping with chirped coherent detection enables single-acquisition retrieval of multi-channel amplitude and phase information without local oscillator source sweeping. Experimental characterization of 16 × 40 Gbit/s QPSK signals verifies a scalable approach for high-capacity microcombenabled transmission systems.
Light Fidelity (LiFi) can provide robust and secure wireless connections with high data rates and low latency, making it suitable for demanding applications in industries, hospitals, and enterprises. The multi-user multiple-input multiple-output (MU-MIMO) scenario has already been investigated for LiFi. Meanwhile, IEEE Std 802.11bb has integrated LiFi into the 802.11 framework, demonstrating how existing Wi-Fi chipsets can be reused. However, 802.11 employs specific MU-MIMO algorithms and their performance for LiFi has not yet been explored. In this paper, we study the use of MIMO algorithms in 802.11 for LiFi in single- and multi-user scenarios by using distributed optical frontends in the infrastructure and angular diversity at mobile stations. We derive the bit error rate (BER) for zero-forcing (ZF) precoding and determine both the spatial transmission mode and the modulation and coding scheme (MCS) that maximize the achievable throughput. Simulation results demonstrate maximum user rates of nearly 2 Gbps comparable to state-of-the-art Wi-Fi access points. The results indicate that LiFi can be used to create “optical hotspots”, increasing the spatial reuse beyond current RF capabilities. Overall, our findings show that conventional MU-MIMO schemes from Wi-Fi can be immediately reused for LiFi and, high performance is achievable by optimizing the deployment.
The paper discusses the implementation and analysis of a digital-twin-orchestrated preemptive-switching scenario for sensor-data transmission over physical network interface links. The experimental setup in this work uses physical interface connections based on radio and optical interfaces, defined by the standards of IEEE 802.11n wireless local area network (WLAN) and IEEE 802.15.7-based visible light communications, respectively. Data transmission at physical interfaces is managed using a software-defined network controller that defines traffic flows based on processing commands from the network digital twin, which is modeled for telemetry analysis and decision making. The statistical average of the measurements shows a delay of less than 200 ms for achieving the switching procedure. The implementation provides benefits for wireless sensor networks and the Internet of Things, improving the availability and resilience of sensor nodes by complementing radio-sensitive environments for critical data transmission.
Optical Camera Communication (OCC) has emerged as a promising branch of Optical Wireless Communications (OWC), leveraging the unlicensed optical spectrum and widespread camera-equipped devices to enable low-cost, secure, and interference-resilient communication. As Internet of Things (IoT) deployments grow, conventional radio-frequency (RF) technologies face challenges related to spectrum congestion and electromagnetic interference. OCC addresses these limitations by offering a complementary solution for IoT applications where moderate data rates can be traded for robustness, spatial selectivity, and seamless integration with existing infrastructure. This review provides a comprehensive overview of OCC technologies, examining both transmitter and receiver architectures. Furthermore, this work identifies technical challenges hindering large-scale adoption and outlines emerging research directions towards sixth-generation (6G) networks. Special emphasis is placed on Integrated Sensing and Communication (ISAC) paradigms, and on distributed intelligent OCC networks.
Video-oculography (VOG) provides a fast, noninvasive way to quantify eye-movement and pupil responses linked to how the brain samples and interprets visual information. We propose a task-free, emotionally valenced viewing paradigm for automatic screening of neurodegenerative disorders. Participants freely viewed a curated set of IAPS (International Affective Picture System) images spanning positive/negative/neutral valence and face/object content while gaze and pupil signals were recorded. From these recordings, we derived compact oculomotor and pupillary descriptors that capture spatial viewing allocation, eye movements, pupil dynamics, and frequency dynamics, and used them to discriminate healthy controls from amnestic mild cognitive impairment (aMCI), behavioral-variant frontotemporal dementia (bvFTD), and posterior cortical atrophy (PCA). Best configurations achieved mean subject-level accuracies of $0.871 \pm 0.085$ for healthy controls (CTR) versus aMCI, 0.817±0.060 for CTR versus bvFTD, and $0.786 \pm 0.060$ for CTR versus PCA, supporting affect-sensitive free-viewing VOG as a low-burden complementary screening signal.
In robotic manipulation tasks localisation and alignment is critical, particularly in indoor environments with inteference from multiple devices. This work presents a markerbased alignment approach for a plastic optical fibre-based visible light communication link applied to a robotic arm. The proposed method combines computer vision and optical wireless communication by using 6x6 ArUco markers for estimating distance and pose respect to an onboard camera mounted on the robotic arm, enabling real-time localisation. In parallel, the system is designed to support VLC through a plastic optical fibre, open in one end, facilitating simultaneous alignment and data transmission.
Data-driven Indoor positioning systems (IPS) commonly require ground truth (GT) labels and are prone to changing environments. This makes deployment labour-intensive and difficult to scale. This work proposes a label-efficient and selfsupervised ultrasonic IPS that removes these requirements by learning the spatial relationships of the environment directly from received channel observations. The method builds on channel charting (CC), using a triplet-based training objective and a graph neural network (GNN) to generate a latent chart that preserves geometric neighbourhood. An affine alignment step maps the learned chart to physical space using only a minimal number of reference points. The system is implemented and evaluated in an indoor testbed. Results demonstrate that the CC-based approach achieves positioning accuracy close to conventional supervised learning while requiring neither GT labels nor anchor coordinate knowledge, highlighting its potential for effortless and scalable deployment. Additional learning during operation mitigates the difficulties arising from a dynamically changing environment.
This paper presents the evaluation for VLC/IR optical wireless architectures for NLoS-only underground tunnels scenarios, modeling the system as a bidirectional link with VLC uplink and IR downlink, evaluating its performance under a load-balancing approach that reallocates traffic according to link availability. Four configurations are considered, corresponding to levels of directional redundancy. The baseline 1-1 configuration is supported by a practical testbed, whereas the remaining cases are examined through system-level comparative extensions under the same traffic-allocation and link-availability, while the system-level performance is evaluated in relation with the load-balance, including throughput, unmet demand, per-link utilization, and fairness. In addition, an illustrative bit error rate (BER) analysis is used to relate physical-layer behavior with system-level robustness. The results show that increasing directional redundancy improves service continuity by mitigating throughput degradation during link DOWN events through traffic reallocation. While configurations with single-link directions exhibit single points of failure, the 2-2 configuration achieves the highest aggregate throughput, the lowest unmet-demand ratios, and the most robust behavior under partial link failures. The BER trends further indicate that parallel link availability reduces sensitivity to individual channel degradation, reinforcing the benefits of balanced uplink and downlink redundancy in tunnel environments.
Accurate estimation of the local number of components is critical in time-frequency (TF) signal analysis, as it underpins reliable detection and preservation of auto-terms. Existing local Rényi entropy (LRE) methods have notable limitations: the non-iterative approach is robust but may miss weak components, whereas the iterative approach may erroneously remove multiple components within a single iteration when components are closely spaced or intersect in the TF domain. This paper presents a TF-filtered iterative LRE framework that isolates components one by one across iterations. The proposed method is integrated into a compressive sensing-based TF reconstruction framework, where local component estimation is used to guide the shrinkage process. Experiments on synthetic signals and real-world electroencephalogram (EEG) seizure data demonstrate that the proposed approach outperforms existing LRE methods in component estimation accuracy and reconstruction quality, while also reducing computational time compared with the conventional iterative LRE. These results highlight the effectiveness of the proposed framework for non-stationary signals exhibiting intersecting TF components.
Whispering-gallery-mode (WGM) resonators offer superior optical confinement and higher resolution for biochemical sensing compared to conventional fiber interferometers. However, traditional functionalization often relies on thick polymer coatings that introduce structural instability and hinder direct light-matter interaction. In this work, we propose a robust, coating-free WGM resonator platform featuring a molecular-level surface functionalization strategy. A key innovation is the “closed-vessel” surface modification process, which protects the fragile sensing region from mechanical damage and contamination. For practical implementation, a specialized 3D-printed sensor holder is developed to enable liquid-phase measurements with minimal analyte volume. A wavelength shift of 0.15549 nm is observed when changing the liquid from deionized water to pH 0.44 hydrochloric acid. Experimental results using a bare-fiber resonator demonstrate excellent reversibility and structural integrity when cycling between deionized water and acid solutions. The proposed platform is easy to fabricate and offers a versatile template for various reliable sensing applications, ranging from environmental monitoring to real-time chemical detection.
This paper presents a three-dimensional (3D) unmanned aerial vehicle (UAV) localization framework based on optical round-trip time (RTT) ranging and multilateration. Multiple fixed optical transmitters estimate distances via laser reflections from a UAV equipped with passive retroreflectors, enabling infrastructure-assisted localization without active onboard hardware or additional energy consumption. The system model incorporates additive Gaussian ranging noise, and the localization problem is formulated as a non-linear least-squares estimation problem. Monte Carlo simulations demonstrate a median localization error of approximately 0.055 m, with 90% and 95% of errors below 0.11 m and 0.14 m, respectively, under centimeter-level ranging noise. The results show that localization accuracy depends on transmitter geometry and UAV altitude, with errors increasing at higher altitudes and under larger noise levels. Furthermore, increasing the number of transmitters improves accuracy through enhanced measurement redundancy. The results provide practical design guidelines for optical UAV tracking systems.
Scalable storage systems typically provide mechanisms for re-distributing data over time to ensure a balanced storage allocation. Such mechanisms have traditionally been considered as background activities meant to run at low priority, to avoid penalizing applications accessing storage. In this work, we focus on an alternative design point where data re-distribution is run as a high-priority activity, meant to re-balance data at the full speed allowed by the newly added resources. Such rapid redistribution has become viable in recent years due to the rise of low-overhead technologies in the networking and disk storage space. A challenge in such a scenario is to schedule data transfer flows to newly added nodes in a way that fully and efficiently utilizes network, CPU, and disk resources in the new nodes, while avoiding overload. To understand the impact of different parameters of the rebalance process, such as the number of simultaneous senders, we develop a queueing network (QN) model of the process used in MongoDB, and describe a preliminary evaluation of performance prediction by the QN model via simulations.
Most existing time synchronization methods are developed for single-path 5G NR channels. They do not provide satisfactory timing performance for multipath channels. One method was proposed for wireless LAN systems. In principle, it can be used for time synchronization for 5 G NR systems. However, an ill-conditioning is encountered in our implementation of that method. Hence, a nullspace-fitting method is proposed for time synchronization in multipath channels, using the first symbol of the synchronization signal/PBCH block (SSB) of 5G NR systems. This method can provide a range of the initial time instants of the first symbol of the SSB. It significantly outperforms some existing methods in time synchronization.
