3D object detection using LiDAR-based point cloud data and deep neural networks is essential in autonomous driving technology. However, deploying state-of-the-art models on edge devices present challenges due to high computational demands and energy consumption. Additionally, single LiDAR setups suffer from blind spots. This paper proposes SC-MII, multiple infrastructure LiDAR-based 3D object detection on edge devices for Split Computing with Multiple Intermediate outputs Integration. In SC-MII, edge devices process local point clouds through the initial DNN layers and send intermediate outputs to an edge server. The server integrates these features and completes inference, reducing both latency and device load while improving privacy. Experimental results on a real-world dataset show a 2.19× speed-up and a 71.6% reduction in edge device processing time, with at most a 1.09% drop in accuracy.

Federated Learning (FL) enables collaborative model training across decentralized clients, enhancing privacy by keeping data local. Yet conventional FL, relying on frequent parameter-sharing, suffers from high communication overhead and limited model heterogeneity. Distillation-based FL approaches address these issues by sharing predictions (softlabels, i.e., normalized probability distributions) instead, but they often involve redundant transmissions across communication rounds, reducing efficiency. We propose SCARLET, a novel framework integrating synchronized soft-label caching and an enhanced Entropy Reduction Aggregation (Enhanced ERA) mechanism. SCARLET minimizes redundant communication by reusing cached soft-labels, achieving up to 50 in communication costs compared to existing methods while maintaining competitive accuracy. Enhanced ERA resolves the fundamental instability of conventional temperature-based aggregation, ensuring robust control and high performance in diverse client scenarios. Experimental evaluations demonstrate that SCARLET consistently outperforms state-of-the-art distillationbased FL methods in terms of accuracy and communication efficiency. The implementation of SCARLET is publicly available athttps://github.com/kitsuyaazuma/SCARLET.

Split computing (SC) is a distributed inference method designed to balance computational load and reduce latency by partitioning a neural network into head and tail networks, deployed on a mobile device and a server, respectively. Due to the limited bandwidth, the data size at the split point must be reduced. For this reason, layers with small channel sizes, called bottlenecks, are often introduced in the model. However, the degree to which the channel size should be reduced depends on the overall structure of the model and the difficulty of the task, so a search is necessary. In this paper, we propose a training method that simultaneously reduces the number of bottleneck channels and optimizes model parameters for SC. This method introduces weights for the bottleneck layer into the loss function of training using cross-entropy. This makes the output value of each channel in the bottleneck layer closer to zero, and reduces the size of the transmitted data by filling in the values close to zero at the server side instead of sending them by the client during inference.

WiFi channel state information (CSI) and machine learning-based environmental sensing have gained significant attention. However, it is empirically known that the accuracy of trained models degrades over time. In this study, we experimentally demonstrate that this temporal degradation in CSI sensing model accuracy is caused by concept drift, a phenomenon where the distribution of CSI shifts over time. Specifically, we provide evidence of long-term concept drift based on CSI measurements taken in a static environment over the course of seven days. Furthermore, we show that by applying noise-based data augmentation during model training, the impact of long-term concept drift can be mitigated. This approach alleviates the need for continuous model fine-tuning, which is resource-intensive due to its reliance on costly labeled data.

This study explores Wi-Fi Channel State Information (CSI) as a novel tactile sensing approach to overcome the limitations of existing methods. Tactile sensing, particularly for detecting object deformation due to human touch, is crucial in human-robot interaction, especially with the rise of collaborative robots and assistive devices. Traditional tactile sensors are effective but face challenges like complex wiring, fragility, and the risk of electric leakage. Vision-based sensing provides accurate deformation detection but is hindered by visibility constraints and high costs. Our research leverages CSI obtained from WiFi signals propagating through an object to estimate its deformation. CSI sensing is an emerging integrated sensing and communication (ISAC) technique that captures environmental changes via wireless signal propagation, offering a cost-effective alternative. We introduce a machine-learning model trained to predict object shape from CSI data, validated against a camera-based shape sensor. Experimental results using IEEE 802.11n WiFi devices, ESP32, and an aluminum pipe demonstrate that our CSI-based approach can accurately predict object deformation, suggesting it as a promising alternative to conventional methods. This research opens new possibilities for tactile sensing in soft robotics and broadens the application range of CSI sensing.

This paper proposes Load-aware Tram-FL, an extension of Tram-FL that introduces a training scheduling mechanism to minimize total training time in decentralized federated learning by accounting for both computational and communication loads. The scheduling problem is formulated as a global optimization task, which—though intractable in its original form—is made solvable by decomposing it into node-wise subproblems. To promote balanced data utilization under non-IID distributions, a variance constraint is introduced, while the overall training latency, including both computation and communication costs, is minimized through the objective function. Simulation results on MNIST and CIFAR-10 demonstrate that Load-aware Tram-FL significantly reduces training time and accelerates convergence compared to baseline methods.

We demonstrate real-time high-resolution generation of images of the physical environment from WiFi CSI. Our demo is based on LatentCSI, our novel CSI-to-image generation framework that encodes CSI samples into the latent space of a pretrained latent diffusion model (LDM) to produce high-quality images with optional editing & reconstruction capability by text prompts. Our prototype consists of sensor nodes to collect ground truth CSI and camera images, an edge and training server to host and update LatentCSI components, and a client server that serves predicted images to clients. Our use of latent space offers faster training and inference at better quality. This allows for online model training at ~30 samples/sec, and inference with a latency of ~60ms. To the best of our knowledge, this work presents the first real-time demonstration of image generation from WiFi CSI.

We introduce CSI-Inpainter, a novel approach for obstacle removal using Wi-Fi CSI. This method harnesses CSI data to reconstruct obscured visual elements, regardless of lighting conditions. Extensive empirical evaluation in both office and industrial settings demonstrates the effectiveness of CSI-Inpainter’s exceptional ability to identify and reconstruct occluded segments, outperforming traditional baselines and our received signal strength indicator (RSSI)-based work, RF-Inpainter in terms of visual quality. Our findings emphasize the superiority of CSI data over RSSI for providing richer visual information and underscore the critical role of optimal sensor placement and data fusion from multiple CSI sensors in enhancing the performance. CSI-Inpainter represents a significant advancement in obstacle removal for various applications like surveillance, offering new insights into the integration of wireless sensing and visual scene recovery, expanding the potential applications of Computer Vision in real-world environments.

In decentralized federated learning (DFL), the dual challenges of extensive inter-node communication and non-independent, identically distributed (non-IID) data impede the attainment of high-accuracy models while maintaining minimal communication traffic. We propose Tram-FL, which progressively refines a global model by transferring it sequentially amongst nodes. We also introduce a dynamic model routing algorithm for optimal route selection, aimed at enhancing model precision with minimal forwarding. Our experiments demonstrate that Tram-FL with the proposed routing delivers high model accuracy, outperforming baselines while reducing communication costs.

Split computing (SC) introduces an innovative framework designed for inference operations of deep neural networks (DNNs), enhancing traditional machine learning inference approaches that typically deploy models solely on local devices or edge/cloud servers. This research not only implements the SC framework for distributed machine learning inference but also showcases its application in computer vision tasks. We demonstrate the capability of SC to execute privacy-preserving distributed inferences across both image classification and object detection tasks. Our comprehensive implementation will be made publicly available at https://github.com/nishio-laboratory/.

This paper demonstrates the feasibility of millimeter-wave (mmWave) link quality prediction based on out-of-band channel state information (CSI) fingerprinting. To overcome the pedestrian blockage problem of mmWave communications, a large number of computer vision-aided mmWave link quality prediction methods have been investigated. However, the use of cameras and LiDAR to acquire computer vision information entails privacy risks. In this paper, we employ 5 GHz band CSI fingerprinting - an aggregation of CSI measured at multiple locations, for mmWave link quality prediction. CSI reflects the propagation environment of the wireless communication channel and thus includes information on pedestrians that block mmWave communications. CSI fingerprinting, aggregated from various measurement locations, enables future mmWave link quality prediction owing to its sufficient spatial information. We conducted a real-world wireless communication experiment with commercial devices compliant with IEEE 802.11ad for the mmWave, and nine IEEE 802.11ac CSI measurement devices for 5 GHz, to experimentally evaluate our method. The experimental result revealed that our proposed method can deterministically and numerically predict the mmWave link quality deterioration caused by pedestrian blockage 500 ms in advance.

Recomposing channel state information (CSI) from the beamforming feedback matrix (BFM), which is a compressed version of CSI and can be captured because of its lack of encryption, is an alternative way of implementing firmware-agnostic WiFi sensing. In this study, we propose the use of camera images toward the accuracy enhancement of CSI recomposition from BFM. The key motivation for this vision-aided CSI recomposition is to draw a first-hand insight that the BFM does not fully involve spatial information to recompose CSI and that this could be compensated by camera images. To leverage the camera images, we use multimodal deep learning. We conducted experiments using IEEE 802.11ac devices and revealed that the recomposition accuracy of the proposed multimodal framework is improved compared to the single-modal framework only using images or BFMs.

This paper presents the watch-from-sky framework, where multiple unmanned aerial vehicles (UAVs) play four roles, i.e., sensing, data forwarding, computing, and patrolling, for predictive police surveillance. This paper reports a simulation of UAV dispatching using reinforcement learning and distributed ML inference.

Asynchronous decentralized federated learning is a promising distributed machine learning framework from the viewpoint of the server cost saving and communication bottleneck mitigation as well as privacy and security protection. Although several methods have been proposed for asynchronous decentralized federated learning, most existing works have a problem of performance degradation for non-independent and identically distributed (non-IID) data. Even the methods aimed for non-IID data do not achieve high accuracy or fast convergence for highly non-IID data in sparse network topologies. To address this issue, we propose a new parameters exchange approach "Skip" and employ it in combination with an existing approach "Swap" to let the distributed models efficiently learn non-local data. Then, we propose two novel asynchronous decentralized federated learning methods, Greedy-Skip & Swap SGD (GSS SGD) and Topology-aware-Skip & Swap SGD (TSS SGD), by combining Skip and Swap in a topology-agnostic and topology-aware fashion, respectively. Our evaluation demonstrated that our TSS SGD outperforms existing methods for highly non-IID data, in terms of the inference accuracy and convergence speed, regardless of the sparsity of topologies.

This paper proposes a feature extraction method from three-dimensional (3D) point cloud employing spatial frequency analysis for point cloud-based link quality prediction. The proposed method aims to address the human blockage problem in millimeter-wave (mmWave) communications, where proactive communication control through machine learning-based future link quality prediction has been shown to be effective in preventing link quality degradation. However, the use of 3D point clouds presents challenges such as increased transmission cost and computational complexity due to the large data size. To address these challenges, we propose a preprocessing method that can efficiently extract features from the point cloud and reduce data size without compromising the accuracy of mmWave link quality prediction. Our approach uses spatial frequency-based filtering to isolate signals related to moving obstacles, such as pedestrians, in the frequency domain. It also removes large, static background objects like walls and furniture. The proposed method can extract relevant objects in the point cloud based on their spatial frequency without requiring object detection or segmentation. The experimental results demonstrate that our proposed method significantly reduces feature data size by approximately 99% compared to conventional methods, while still maintaining high link quality prediction accuracy.

This study develops a federated learning (FL) framework overcoming largely incremental communication costs due to model sizes in typical frameworks without compromising model performance. To this end, based on the idea of leveraging an unlabeled open dataset, we propose a distillation-based semi-supervised FL (DS-FL) algorithm that exchanges the outputs of local models among mobile devices, instead of model parameter exchange employed by the typical frameworks. In DS-FL, the communication cost depends only on the output dimensions of the models and does not scale up according to the model size. The exchanged model outputs are used to label each sample of the open dataset, which creates an additionally labeled dataset. Based on the new dataset, local models are further trained, and model performance is enhanced owing to the data augmentation effect. We further highlight that in DS-FL, the heterogeneity of the devices’ dataset leads to ambiguous of each data sample and lowing of the training convergence. To prevent this, we propose entropy reduction averaging, where the aggregated model outputs are intentionally sharpened. Moreover, extensive experiments show that DS-FL reduces communication costs up to 99 percent relative to those of the FL benchmark while achieving similar or higher classification accuracy.

With regard to the implementation of WiFi sensing agnostic according to the availability of channel state information (CSI), we investigate the possibility of estimating a CSI matrix based on its compressed version, which is known as beamforming feedback matrix (BFM). Being different from the CSI matrix that is processed and discarded in physical layer components, the BFM can be captured using a medium-access-layer frame-capturing technique because this is exchanged among an access point (AP) and stations (STAs) over the air. This indicates that WiFi sensing that leverages the BFM matrix is more practical to implement using the pre-installed APs. However, the ability of BFM-based sensing has been evaluated in a few tasks, and more general insights into its performance should be provided. To fill this gap, we propose a CSI estimation method based on BFM, approximating the estimation function with a machine learning model. In addition, to improve the estimation accuracy, we leverage the inter-subcarrier dependency using the BFMs at multiple subcarriers in orthogonal frequency division multiplexing transmissions. Our simulation evaluation reveals that the estimated CSI matches the ground-truth amplitude. Moreover, compared to CSI estimation at each individual subcarrier, the effect of the BFMs at multiple subcarriers on the CSI estimation accuracy is validated.

In IEEE 802.11bc, the broadcast mode on wireless local area networks (WLANs), data rate control that is based on acknowledgement (ACK) mechanism similar to the one in the current IEEE 802.11 WLANs is not applicable because the ACK mechanism is not implemented. This paper addresses this challenge by proposing ACK-less data rate adaptation methods by capturing non-broadcast uplink frames of STAs. In IEEE 802.11bc, a use case is assumed, where a part of STAs in the broadcast recipients is also associated with non-broadcast APs, and such STAs periodically transmit uplink frames including ACK frames. The proposed method is based on the idea that by overhearing such uplink frames, the broadcast AP surveys channel conditions at partial STAs, thereby setting appropriate data rates for the STAs. Furthermore, to avoid reception failures in a large portion of STAs, this paper proposes deep reinforcement learning (DRL)-based data rate adaptation framework that uses a sim-to-real approach. Therein, information of reception success/failure at broadcast recipient STAs, that could not be notified to the broadcast AP in real deployments, is made available by simulations beforehand, thereby forming data rate adaptation strategies. Numerical results show that utilizing overheard uplink frames of recipients makes it feasible to manage data rates in ACK-less broadcast WLANs, and using the sim-to-real DRL framework can decrease reception failures.

To mitigate cyberattacks, detecting anomalies in network traffic is of key importance. In this paper, we propose a model training method for detection of Internet of Things (IoT) anomalous traffic that is robust against the contamination of anomalous samples in the training set. The key idea is to focus on the nature of IoT malware infections (i.e., only a limited number of IoT networks contain infected devices) and employ federated learning (FL) to mitigate the impact of anomalous samples on model training. The simulation evaluation using IoT traffic data obtained from residences and malware traffic data collected from sandbox experiments demonstrates that the proposed method does not cause accuracy degradation even when the anomalous samples are contaminated, in contrast with the detection accuracy of baseline methods, which does degrade.

Underwater networks are expected to be service platforms for broad-sea and deep-sea activities. The significant challenge of underwater communication has been achieving high-speed and long-distance data transmission due to the high-attenuation and time-varying channel state in underwater environments. It is reasonable to get the underwater data above the water surface for establishing underwater-ground networks. However, it is still an unsolved issue to efficiently establish underwater-ground communication channel. To address this problem, we propose an aquatic fronthaul for underwater-ground communication, where floating aquatic relay nodes relay data from underwater drones/sensors to a ground radio unit. We propose a relocation algorithm for aquatic relay nodes to efficiently reconstruct the network according to the distribution of underwater nodes. The advantage of the proposed algorithm is robustness for the uncertainty of underwater node locations due to the difficulty in underwater localization. The performance of the proposed algorithm was evaluated with multi-agent simulations. The feasibility of the aquatic fronthaul network was confirmed via the experimental results with a Wi-Fi mesh network above the water.

This paper demonstrates the feasibility of predicting the future received power of millimeter-wave (mmWave) communication using point cloud data. To mitigate the human blockage problem in mmWave communication, previous works have studied a camera-vision assisted mmWave link quality prediction, which predicts the time-series of the received power from the next moment to as many as several hundred milliseconds ahead by leveraging camera imagery and machine learning. However, camera imagery generally includes privacy-sensitive information, which induces privacy concerns in the camera-assisted mmWave networks. In this paper, we demonstrate that point cloud, which can be obtained by light detection and ranging (LiDAR) sensors and poses fewer privacy concerns than camera imagery, can be an alternative to the cameras in mmWave link quality prediction. Specifically, we propose a mmWave received power prediction method using point cloud data, and our experimental evaluation demonstrates that the proposed method predicts mmWave received power 500ms ahead with a root-mean-squared error of 3.3dB, which is comparable to the existing camera-based method. Moreover, we verify that even minor environmental changes can degrade the accuracy of the trained prediction model and this degradation can be mitigated by model fine-tuning with a small dataset.

Distributed inference is an emerging technology that enables inference with cutting-edge machine learning (ML) models such as deep neural networks (DNNs) on resource-constrained devices. However, narrow-band and lossy wireless networks easily create bottlenecks and increase the latency in distributed inference. This study proposes the joint optimization of an ML model and wireless communication parameters (such as transmission rate and retransmission limit) to reduce communication latency while maintaining the accuracy of inference. Our key idea is to utilize the packet-loss tolerance of ML inference that decreased the reliability but reduced communication latency. To this end, the proposed method based on the multi-armed bandit (MAB) algorithm, namely the upper confidence bound (UCB) algorithm, jointly optimizes (i) the wireless communication parameters that control the trade-off between reliability and latency in communications and (ii) the architecture of the ML model that controls the trade-off between accuracy and packet-loss reliance in ML inference. The results of computer simulations using ns3-ai show that the proposed method of joint optimization maintains the accuracy of inference as well as achieves a lower latency than when only the architecture of the ML model or communication parameters are optimized.

The IEEE 802.11ax has introduced the concurrent transmissions among neighboring wireless local area networks (WLANs) to facilitate spatial reuse. As a side effect, the scheduling problem has become more challenging due to the appearance of unmanaged co-channel interference. This paper proposes a latency-aware fair resource scheduling scheme in dense WLANs with Lyapunov optimization. In the scheduling scheme, we formulate a scheduling problem for WLANs with the stabilization of the transmission queue, which is one of the quality of service (QoS) issues and the fairness of the instantaneous data rate as a stochastic optimization problem. This problem is solved by using Lyapunov optimization, which eliminates the time average constraints of the problem. We performed numerical simulations in dense WLANs, and the results confirm that the proposed scheme guarantees an allowable queue size. the results also show that the proposed scheme achieves a higher fairness index and a smaller queue size than compared methods.

A key aspect to realize Internet of things applications such as industry automation and smart agriculture is to enable realtime and networked automatic monitoring via cloud computing and computer vision. However, to design a networked monitoring system, it is necessary to realize a balance between the monitoring accuracy and monitoring cost, for instance, between the network traffic to transmit images and the computation load. Although the monitoring cost can be decreased by increasing the sampling interval of cameras, it becomes more likely that informative images cannot be obtained; in other words, the monitoring accuracy decreases with a reduction in the amount of data. Moreover, although on-device image processing can decrease the network traffic, a large computation delay may be incurred, limiting the sampling rate of the monitoring system. The objective of this study was to examine the balance between the monitoring accuracy and cost and to develop a joint optimization technique for the sampling interval and computation offloading to minimize the monitoring cost in a networked monitoring system while ensuring a high monitoring accuracy. The main contributions of this paper are that we prove the joint optimization problem can be solved explicitly and to develop an algorithm to obtain the solution of the joint optimization problem. The simulation results demonstrated that the proposed algorithm can reduce the monitoring cost by 24-48% while maximizing the number of nodes ensured to achieve high monitoring accuracy.

The distributed inference framework is an emerging technology for real-time applications empowered by cutting-edge deep machine learning (ML) on resource-constrained Internet of things (IoT) devices. In distributed inference, computational tasks are offloaded from the IoT device to other devices or the edge server via lossy IoT networks. However, narrow-band and lossy IoT networks cause non-negligible packet losses and re-transmissions, resulting in non-negligible communication latency. This study solves the problem of the incremental retransmission latency caused by packet loss in a lossy IoT network. We propose a split inference with no retransmissions (SI-NR) method that achieves high accuracy without any retransmissions, even when packet loss occurs. In SI-NR, the key idea is to train the ML model by emulating the packet loss by a dropout method, which randomly drops the output of hidden units in a neural network layer. This enables the SI-NR system to obtain robustness against packet losses. Our ML experimental evaluation reveals that SI-NR obtains accurate predictions without packet retransmission at a packet loss rate of 60%.

We propose a sophisticated network selection scheme based on multi-armed bandits and stochastic geometry for heterogeneous cellular networks. In the system model, a user seeking the best network tries to estimate the density of active interferers for every network through the repeated observation of signal-to-interference power ratio (SIR), which shows the randomness induced by randomized interference sources and fading effects. The purpose of this study is to enable the user to identify the network with the lowest density of active interferers while considering the communication quality during exploration. In order to resolve the trade-off between getting more observations on uncertain networks and using a network that seems better so far, we employ a bandit algorithm called Thompson sampling (TS), which is known for its empirical effectiveness. We take two ideas into consideration to enhance TS. First, noticing that the statistical SIR model given by stochastic geometry is useful for capturing the relationship between observed SIR and density of active interferers, we propose to incorporate the statistical model into TS. Second, TS requires us to sample from the posterior distribution of the density parameter for each network, while the distribution obtained through stochastic geometry is much more complicated to generate samples than well-known distribution; we reveal that such a sampling process is achieved with the help of the Markov chain Monte Carlo method. The simulation results show that the proposed method enables a user to find the best network more efficiently than well-known bandit algorithms such as an ϵ-greedy strategy.

For reliable millimeter-wave (mmWave) networks, this paper proposes cooperative sensing with multi-camera operation in an image-to-decision proactive handover framework that directly maps images to a handover decision. In the framework, camera images are utilized to allow for the prediction of blockage effects in a mmWave link, whereby a network controller triggers a handover in a proactive fashion. Furthermore, direct mapping allows for the scalability of the number of pedestrians. This paper experimentally investigates the feasibility of adopting cooperative sensing with multiple cameras that can compensate for one another's blind spots. The optimal mapping is learned via deep reinforcement learning to resolve the high dimensionality of images from multiple cameras. An evaluation based on experimentally obtained images and received powers verifies that a mapping that enhances channel capacity can be learned in a multi-camera operation. The results indicate that our proposed framework with multi-camera operation outperforms a conventional framework with single-camera operation in terms of the average capacity.

We propose a sequential channel allocation method based on the multi-objective multi-armed bandit (MOMAB) problem to identify the best channel set among model-based solutions. A solution obtained from pre-designed objective functions cannot always be the best channel set due to external factors in the wireless environment. It is difficult to take into account external factors in advance, so we propose to allocate channels based on several performance metrics that can only be measured by operating access points. The fine-tuning during actual operation involves a trade-off between exploration and exploitation for the best channel set. In addition, we should utilize a channel set that performs not so good on some metrics but well on other metrics. By using MOMAB, we can balance between exploration and exploitation of Pareto optimal channel sets for multiple metrics. The experimental results demonstrate that the proposed method successfully identifies a Pareto optimal channel set.

In this study, we extended a contextual bandit algorithm, LinUCB, to facilitate cooperative learning of an optimal strategy with intermittent information sharing. We then applied the algorithm to a Wi-Fi basic service set (BSS) selection problem. The BSS selection problem, in which a mobile user selects a BSS that provides maximal throughput based on information observed, still remains a topic of debate. Reinforcement learning, specifically the multi-armed bandit algorithm, enables mobile users to learn an optimal strategy for selecting a good BSS in their environments. This paper proposes a cooperative contextual bandit algorithm, called Cooperative LinUCB (CoopLinUCB), to address the BSS selection problem. Conventional cooperative bandit algorithms require to share experiences such as context and payoffs for every action, and the information sharing increases communication costs. The proposed algorithm enables mobile users to learn strategies using a limited amount of information that is shared intermittently. The learned strategies are then guaranteed to be equivalent to the strategies that are updated using user experience information. Simulation evaluation based on measured throughput and received signal strength indication fingerprint demonstrates that CoopLinUCB-based BSS selection learns a BSS selection strategy faster and reduces the cumulative regret compared to BSS selection without cooperation.

Towards ultra-reliable and low-latency communications, this paper proposes a packet mapping algorithm in an enhanced distributed channel access (EDCA) scheme using policy gradient reinforcement learning (RL). The EDCA scheme provides higher priority packets with more transmission opportunities by mapping packets to a predefined access category (AC); thereby, the EDCA scheme supports a higher quality of service in wireless local area networks. In this paper, it is noted that by mapping high priority packets to lower priority ACs, the one-packet delay of a high priority packet can be reduced. In contrast, the mapping algorithm cannot minimize the multiple-packets delay because the mapping algorithm is based on the current status. This is because, from a long-term perspective, mapping high priority packets is required as a countermeasure for collisions, to minimize the multiple-packets delay. As a solution, this paper proposes a new mapping algorithm using RL because RL is suitable for maximizing the reward from a long-term perspective. The key idea is to design the state such that the state involves the number of packets having arrived at each AP in the past, which is an indicator expressing past status. In the designed RL task, the reward, i.e., the multiple-packets delay depends on an overall sequence of states and actions; hence, the recursive value function-based RL algorithms are not compatible. To solve this problem, this paper utilizes policy gradient RL, which learns the packet mapping policy from an overall state-action sequence and a consequent multiple-packets delay. The simulation result reveals that the transmission delay of the proposed mapping algorithm is 13.8% shorter than that of the conventional EDCA mapping algorithm.

Over-the-air computation (AirComp)-based federated learning (FL) enables low-latency uploads and the aggregation of machine learning models by exploiting simultaneous co-channel transmission and the resultant waveform superposition. This study aims at realizing secure AirComp-based FL against various privacy attacks where malicious central servers infer clients' private data from aggregated global models. To this end, a differentially private AirComp-based FL is designed in this study, where the key idea is to harness receiver noise perturbation injected to aggregated global models inherently, thereby preventing the inference of clients' private data. However, the variance of the inherent receiver noise is often uncontrollable, which renders the process of injecting an appropriate noise perturbation to achieve a desired privacy level quite challenging. Hence, this study designs transmit power control across clients, wherein the received signal level is adjusted intentionally to control the noise perturbation levels effectively, thereby achieving the desired privacy level. It is observed that a higher privacy level requires lower transmit power, which indicates the tradeoff between the privacy level and signal-to-noise ratio (SNR). To understand this tradeoff more fully, the closed-form expressions of SNR (with respect to the privacy level) are derived, and the tradeoff is analytically demonstrated. The analytical results also demonstrate that among the configurable parameters, the number of participating clients is a key parameter that enhances the received SNR under the aforementioned tradeoff. The analytical results are validated through numerical evaluations.

This paper proposes a cooperative mechanism for mitigating the performance degradation due to non-independent and-identically-distributed (non-IID) data in collaborative machine learning (ML), namely federated learning (FL), which trains an ML model using the rich data and computational resources of mobile clients without gathering their data to central systems. The data of mobile clients is typically non-IID owing to diversity among mobile clients' interests and usage, and FL with non-IID data could degrade the model performance. Therefore, to mitigate the degradation induced by non-IID data, we assume that a limited number (e.g., less than 1%) of clients allow their data to be uploaded to a server, and we propose a hybrid learning mechanism referred to as Hybrid-FL, wherein the server updates the model using the data gathered from the clients and aggregates the model with the models trained by clients. The HybridFL solves both client- and data-selection problems via heuristic algorithms, which try to select the optimal sets of clients who train models with their own data, clients who upload their data to the server, and data uploaded to the server. The algorithms increase the number of clients participating in FL and make more data gather in the server IID, thereby improving the prediction accuracy of the aggregated model. Evaluations, which consist of network simulations and ML experiments, demonstrate that the proposed scheme achieves a 13.5% higher classification accuracy than those of the previously proposed schemes for the non-IID case.

Federated learning (FL) enables a neural network (NN) to be trained using privacy-sensitive data on mobile devices while retaining all the data on their local storages. However, FL asks the mobile devices to perform heavy communication and computation tasks, i.e., devices are requested to upload and download large-volume NN models and train them. This paper proposes a novel unsupervised pre-training method adapted for FL, which aims to reduce both the communication and computation costs through model compression. Since the communication and computation costs are highly dependent on the volume of NN models, reducing the volume without decreasing model performance can reduce these costs. The proposed pretraining method leverages unlabeled data, which is expected to be obtained from the Internet or data repository much more easily than labeled data. The key idea of the proposed method is to obtain a "good" subnetwork from the original NN using the unlabeled data based on the lottery hypothesis. The proposed method trains an original model using a denoising auto encoder with the unlabeled data and then prunes small-magnitude parameters of the original model to generate a small but good subnetwork. The proposed method is evaluated using an image classification task. The results show that the proposed method requires 35% less traffic and computation time than previous methods when achieving a certain test accuracy.

This paper proposes a method to predict received power in urban area deterministically, which can learn a prediction model from small amount of measurement data by a simulation-aided transfer learning and data augmentation. Recent development in machine learning such as artificial neural network (ANN) enables us to predict radio propagation and path loss accurately. However, training a high-performance ANN model requires a significant number of data, which are difficult to obtain in real environments. The main motivation for this work was to facilitate accurate prediction using small amount of measurement data. To this end, we propose a transfer learning-based prediction method with data augmentation. The proposed method pre-trains a prediction model using data generated from ray-tracing simulations, increases the number of data using simulation-assisted data augmentation, and then fine-tunes a model using the augmented data to fit the target environment. Experiments using Wi-Fi devices were conducted, and the results demonstrate that the proposed method predicts received power with 50% (or less) of the RMS error of conventional methods.

This paper proposes a robust adversarial reinforcement learning (RARL)-based multi-access point (AP) coordination method that is robust even against unexpected decentralized operations of uncoordinated APs. Multi-AP coordination is a promising technique towards IEEE 802.11be, and there are studies that use RL for multi-AP coordination. Indeed, a simple RL-based multi-AP coordination method diminishes the collision probability among the APs; therefore, the method is a promising approach to improve time-resource efficiency. However, this method is vulnerable to frame transmissions of uncoordinated APs that are less aware of frame transmissions of other coordinated APs. To help the central agent experience even such unexpected frame transmissions, in addition to the central agent, the proposed method also competitively trains an adversarial AP that disturbs coordinated APs by causing frame collisions intensively. Besides, we propose to exploit a history of frame losses of a coordinated AP to promote reasonable competition between the central agent and adversarial AP. The simulation results indicate that the proposed method can avoid uncoordinated interference and thereby improve the minimum sum of the throughputs in the system compared to not considering the uncoordinated AP.

The concept of vehicle-mounted crowdsourced radio units (CRUs) for a smart city has been proposed to utilize the power of citizens in the deployment of small cells of the centralized radio access network (C-RAN) architecture. Wireless relay fronthaul networking is a promising solution for efficient utilization of vehicle-mounted small cells. However, there have been no routing schemes that can satisfy the strict delay requirements of mobile fronthaul coping with the high dynamicity of vehicles. Thus, this paper proposes a real-time routing scheme for establishing wireless relay fronthaul with vehicle-mounted CRUs. The route optimization is formulated as a boolean satisfiability problem (SAT), and an FPGA-based SAT solver is employed for the fast computation. It can dynamically optimize the forwarding paths in real-time with the constraints of delay requirements. The performance of the proposed routing scheme is confirmed via computer simulations.

To achieve reliable small cell millimeter-wave wireless backhauls, this study installs small cell base stations (SBSs) on overhead messenger wires to gain flexibility in physical deployments of SBSs ensuring in the line-of-sight connections between SBSs and gateway BSs. These installations pose challenges in aligning directional beams, whereby complicated wind-forced dynamics in on-wire SBSs require frequent beam training, and consequently, a large signaling overhead. To address this, this study aims at demonstrating the feasibility of learning-based beam tracking where a beam tracking policy is learned a priori to fix beam misalignment caused by the wind-forced dynamics. Because wind-forced dynamics in SBSs can be three-dimensional (3D), the proposed beam tracking newly exploits the 3D position/velocity of the SBS as state information. As a solution to fix beam misalignment, the beam tracking policy is learned via deep reinforcement learning wherein the 3D information and beam direction are regarded as a state and an action, respectively, and the received signal power at a gateway BS is maximized. The simulation results depict the feasibility of learning an appropriate beam tracking policy to prevent beam misalignment induced by wind-forced 3D dynamics in on-wire SBSs.

Despite the widespread popularity of stochastic geometry analysis for cellular networks, most analytical results lack the perspective of channel-adaptive user scheduling. This study presents a stochastic geometry analysis of the SINR distribution and scheduling gain of normalized SNR-based scheduling in an uplink Poisson cellular network, in which the per-user truncated fractional transmit power control is performed. Because the effects of multi-user diversity depend on the number of candidate users to be scheduled, which is a random variable in a Poisson cellular network, the number distribution of candidate users is a major factor in analyzing the SINR distribution of user scheduling. However, the maximum transmit power constraint of users complicates the distribution of candidate users. This study provides the number distribution of candidate users in a general form, which is obtained by modeling the area of the existing range of candidate users using a beta distribution. Based on this result, this study successfully obtains the uplink SINR distribution under channel-adaptive user scheduling, including cases in which edge users are both allowed and not allowed to transmit at the maximum transmit power. Numerical evaluations reveal that the scheduling gain varies depending on the SNR and the fraction of edge users.

In this study, we conducted an analysis of the system performance of a wireless local area network in which access points (APs) dynamically adjust the carrier sense threshold (CST) based on the individual average received power to determine the optimal CST. Adjustment of the CST is a promising approach to improve spatial reuse and proposed for IEEE 802.11ax standard. Here, assuming to adopt the inversely proportional setting of the CST and transmission power, we can make the carrier sensing relationship symmetric, restraining throughput starvation. This paper analytically derives the density of successful transmissions (DST) on the basis of stochastic geometry. The DST is a system performance metric which expresses the number of APs whose transmission is successful based on signal-to-interference-plus-noise power ratio. We show that both results of the analytically derived DST and Monte Carlo simulation have the same trend. From the perspective of the derived DST, the optimal parameter setting is also discussed.

This paper proposes a camera-based geo-fencing system for wireless local area networks (WLANs) which enables geo-location based wireless access control to intuitively manage the area where the WLANs are available. The proposed system leverages camera to localize WLAN users accurately and estimates the proximity of users to objects in the real world. Meanwhile, conventional geo-location based access control suffers from low accuracy of RSSI based localization. As an example of geo-location based access control, we execute a WLAN activation control which allows STAs to pre-activate WLAN and associates with access points (APs) so that the power consumption and time to obtain contents are reduced. Experimental results show the feasibility of camera-based geo-fencing.

This paper proposes a design for a primary exclusive region (PER) based on a cylindrical grid model in unmanned aerial vehicle (UAV) networks. UAV communications require using additional frequency bands shared with other systems. When they use these bands, their communications must not interfere with the communications of the primary users (e.g., radar systems). To avoid this interference, a PER should be designed. This paper proposes a complex-shaped PER design based on the radar's antenna pattern to maximize the number of transmitting UAVs, and presents a stochastic geometry analysis of interference in UAV networks. On the assumption that the distribution of UAVs in each grid follows an inhomogeneous Poisson point process, the radar's outage probability is derived. From this analysis, an optimization problem of PER is formulated to maximize the number of transmitting UAVs. Subsequently, the solution of this problem is numerically evaluated for a keyhole antenna model. The results show that the complex-shaped PER is designed corresponding to the radar's antenna pattern and the number of transmitting UAVs increases with an increase in the number of grid divisions.

We envision a mobile edge computing (MEC) framework for machine learning (ML) technologies, which leverages distributed client data and computation resources for training high-performance ML models while preserving client privacy. Toward this future goal, this work aims to extend Federated Learning (FL), a decentralized learning framework that enables privacy-preserving training of models, to work with heterogeneous clients in a practical cellular network. The FL protocol iteratively asks random clients to download a trainable model from a server, update it with own data, and upload the updated model to the server, while asking the server to aggregate multiple client updates to further improve the model. While clients in this protocol are free from disclosing own private data, the overall training process can become inefficient when some clients are with limited computational resources (i.e., requiring longer update time) or under poor wireless channel conditions (longer upload time). Our new FL protocol, which we refer to as FedCS, mitigates this problem and performs FL efficiently while actively managing clients based on their resource conditions. Specifically, FedCS solves a client selection problem with resource constraints, which allows the server to aggregate as many client updates as possible and to accelerate performance improvement in ML models. We conducted an experimental evaluation using publicly-available large-scale image datasets to train deep neural networks on MEC environment simulations. The experimental results show that FedCS is able to complete its training process in a significantly shorter time compared to the original FL protocol.

This paper proposes a novel channel allocation scheme based on the replica exchange Monte Carlo method (REMCMC). Some distributed channel allocation schemes in the literature formulate the channel allocation problem as a potential game, in which the unilateral improvement dynamics is guaranteed to converge to a Nash equilibrium. In general, spatial adaptive play (SAP), which is one of the representative learning algorithms in the potential game-based approach, can reach an optimal Nash equilibrium stochastically. However, this is inefficient for the channel allocation and SAP tends to be stuck in a sub-optimal Nash equilibrium in a limited time. To assist in finding the optimal Nash equilibrium for this kind of channel allocation problem, we apply the REMCMC to the existing potential game-based channel allocation. We show that SAP can be considered as a sampling process of the Boltzmann- Gibbs distribution and sampling methods can be utilized. We evaluated the proposed algorithm through simulations and the results show that the proposed algorithm can find the optimal Nash equilibrium quickly.

This paper proposes a pre-training method for a deep- neural-network (DNN) based received power prediction scheme leveraging transfer learning for millimeter-wave (mmWave) networks. The received power prediction scheme has been proposed for proactive network control, which accurately predicts the received power 500 ms ahead using depth- camera images and a DNN. However, the prediction scheme requires a large number of training datasets and computational resources to prepare the accurate prediction model. In this paper, we propose a pre-training method that reduces the preparation time by leveraging the use of 3D model simulations with signal propagation simulations and transfer learning. The proposed method generates a dataset for pre- training using computer simulations, and trains a prediction model. The pre-trained model is transferred and fine-tuned by using a dataset obtained in a place where the system is actually used so that the model fits to the place. The experimental results show that the computational time of the proposed scheme with an RMS error of less than 5 dB is reduced by 78% compared with the previous work when using the dataset obtained in 60 s.

The centralized radio access network (C-RAN) architecture prevails in beyond 5G mobile networks. Along with the cell size reduction, the efficiency of C-RAN architecture is drastically deteriorated by the spatio-temporal fluctuations in mobile traffic demand. To address this problem, we proposed a concept of adaptive C-RAN architecture for smart cities with crowdsourced radio units (CRUs). The advantages of the proposed scheme are high flexibility and low cost, because the distribution of CRUs follows that of mobile users. This paper introduces the experimental results on the performance of radio units mounted on parked vehicles to show the efficacy of the proposed scheme.

Starvation problems in a dense IEEE 802.11 wireless local area network (WLAN) seriously degrade fairness among links in the network. Some transmitters enjoy the full opportunities to access their channels, while others have very few opportunities to transmit, especially when network traffic is saturated. This paper focuses on the joint channel selection and spatial reuse problems in IEEE 802.11ax WLANs. Our main objective is to improve throughput while mitigating starvation. We formulate a non-cooperative game and proved it to be an exact potential game (EPG). We design a joint channel selection and spatial reuse algorithm with only local information exchange, based on the proposed game model. Convergence of the proposed algorithm is guaranteed by the property of potential games when unilateral improvement dynamics is used as a learning algorithm.

This study demonstrates the feasibility of proactive received power prediction by leveraging spatiotemporal visual sensing information towards reliable millimeter-wave (mmWave) networks. As the received power on a mmWave link can attenuate aperiodically owing to human blockages, a long-term series of the future received power cannot be predicted by analyzing the received signals prior to the blockage occurring. We propose a novel mechanism that predicts the time series of received power from the next moment to as many as several hundred milliseconds ahead. The key idea is to leverage camera imagery and machine learning (ML). Time-sequential images may involve the spatial geometry and mobility of obstacles representing mmWave signal propagation. ML is used to construct a prediction model from a dataset of sequential images labeled with received power in several hundred milliseconds ahead of the time at which each image is obtained. The simulation and experimental evaluations conducted using IEEE 802.11ad devices and a depth camera demonstrated that the proposed mechanism employing convolutional long short-term memory predicted a time series of received power up to 500 ms ahead, with an inference time of less than 3 ms and a root-mean-square error of 3.4 dB.

Camera-assisted millimeter-wave (mmWave) network is a new paradigm for mmWave communications where mobility of obstacles is captured by using RGB and depth cameras and conducts network operations by considering the captured information. For camera-assisted mmWave networks, this paper proposes a recurrent neural network (RNN)-based received signal strength (RSS) estimation scheme using depth camera images. This scheme enables us to estimate the RSS of any mmWave links, including links where nodes are not transmitting frames. An RNN enables us to model the relationship between current RSS and an image time series, which includes information regarding the mobility of nodes and obstacles. Simulation results demonstrate that the RNN-based estimation scheme achieves higher accuracy than that of a multi-layer perceptron.

This paper provides the signal-to-interference-plus-noise ratio (SINR) complimentary cumulative distribution function (CCDF) and average data rate of the normalized SNR-based scheduling in an uplink cellular network using stochastic geometry. The uplink analysis is essentially different from the downlink analysis in that the per-user transmit power control is performed and that the interferers are composed of at most one transmitting user in each cell other than the target cell. In addition, as the effect of multi-user diversity varies from cell to cell depending on the number of users involved in the scheduling, the distribution of the number of users is required to obtain the averaged performance of the scheduling. This paper derives the SINR CCDF relative to the typical scheduled user by focusing on two incompatible cases, where the scheduler selects a user from all the users in the corresponding Voronoi cell or does not select users near cell edges. In each case, the SINR CCDF is marginalized over the distribution of the number of users involved in the scheduling, which is asymptotically correct if the BS density is sufficiently large or small. Through the simulations, the accuracies of the analytical results are validated for both cases, and the scheduling gains are evaluated to confirm the multi-user diversity gain.

This paper provides a bargaining game-based band- width allocation scheme in multi-operator shared millimeter-wave (mmWave) radio access network (RAN). We consider mobile network operators (MNOs) that mutually share their mmWave base stations (BSs) to expand coverage such that the subscribers of one MNO can be associated with the mmWave BSs of other MNOs. Since MNOs are also competitive in nature, there is a necessity for MNOs to negotiate the amount of bandwidth to be allocated to the users of each other. We first evaluate how the amount of allocated bandwidth enhances the success probability both theoretically and through simulations. Then, by using the evolutionary game theory to model the subscription of users, the feasible region of bandwidth is mapped to the feasible region of market state, enabling the MNOs to negotiate based the evolution of the market state. The Nash bargaining solution yields an allocation scheme that maximizes the product of MNOs' increments in the market share.

This paper presents the stochastic geometry analysis and design of a primary exclusive region (PER) for cognitive radio vehicular ad hoc networks (CR-VANETs). Recently, to satisfy the increasing demands for vehicular communications, cognitive radio (CR) technology has been applied to broaden the vehicular communication spectrum. However, while utilizing the licensed spectrum opportunistically as secondary transmitters (STs), vehicles must avoid harmful interference with primary receivers (PRs), e.g., TV broadcasters. In contrast to cooperative spectrum sensing and stand-alone spectrum sensing in most existing works, in this paper, we proposed a geo-location-database-driven opportunistic spectrum access (OSA) approach to maximize the spectrum opportunity of SUs. For the proposed OSA approach, this paper presents a line segment model for CR-VANETs and introduces the allowable transmission probability of STs in each line segment. We theoretically analyze the outage probability of the primary user using stochastic geometry. From these analyses, we introduce the optimization problem to maximize the spectrum opportunity of STs while ensuring a minimal outage threshold for the PR. Subsequently, we numerically evaluate the solution of the optimization problem. Our results show that a PER for CR-VANETs has been successfully designed. With the decrease in the length of the line segment and further detailed information, a smaller and more complex-shaped PER can be achieved.

This paper experimentally finds the optimum number of input images of a machine learning-based mmWave received signal strength (RSS) value prediction scheme from depth images. By modeling the relationships between time-sequential depth images and RSS values based on machine learning, it is possible to predict the future RSS values, and thereby, a predictive handover makes a moment of degradation of the RSS value avoidable. As prediction models of RSS value, three machine learning models are compared: the convolutional neural networks (CNN), the combination of CNN and convolutional long short-term memory (CNN+ConvLSTM), and random forest. As the number of input images increases, the prediction accuracy generally improves, however, too numerous input images may make the prediction accuracy worse because of over-fitting. Experimental results reveal that the number of input images that are input in order to predict the RSS value the most accurately is 16.

In millimeter wave (mmWave) vehicle-toinfrastructure (V2I) communications for autonomous vehicles, the small coverage of road side units (RSUs) is an open problem. We propose a multi-hop relaying method using dynamic vehicle deployment in order to increase the coverage of RSUs. The key idea of our method is to leverage the movement controllability of autonomous vehicles to extend the multi- hop relay distance. The proposed deployment method considers blockage because it is a crucial problem in mmWave communications, although it is not a crucial problem in microwave communications. We formulate the deployment problem as an optimization problem and obtain its lower and upper bounds performances. We also introduce a mmWave connectivity graph, from which the vehicle position that achieves the lower bound performance can be obtained by solving a shortest-path problem. Simulation results demonstrate that even when only 7.5% of all vehicles'' positions are controllable, the proposed deployment method can achieve a coverage of 80%, which is more than twice the coverage achieved by the relaying without deployment.

Cooperative perceptions for autonomous vehicles by sharing image sensor data enhance traffic security. For sharing a large amount of sensor data, millimeter-wave (mmWave) communication is expected to be an enabler of high-throughput communications because of its wide band width and its efficiency in spatial reuse. This paper proposes a concurrent scheduling for sensor data dissemination in mmWave vehicular networks at an intersection. The proposed algorithm improves the region covered by shared data in situations where dissemination time is limited and not all data are disseminated. Improvement is realized by prioritizing data to be forwarded considering the geographical information. The priority control enlarges the average of the area of covered region by 15% at maximum. Meanwhile, it is proved that when sufficient time is available for dissemination, the proposed algorithm guarantees that all vehicles can share their data with each other.

This paper discusses a measurement method of time‐variant attenuation of IEEE 802.11ad wireless LAN signals in the 60 GHz band induced by human blockage. The IEEE 802.11ad access point (AP) transmits frames intermittently, not continuously. Thus, to obtain the time‐varying signal attenuation, it is required to estimate the duration in which the AP transmitted signals. To estimate whether the AP transmitted signals or not at each sampling point, this paper applies a simple two‐state hidden Markov model. In addition, the validity of the model is tested based on Bayesian information criterion in order to prevent model overfitting and consequent invalid results. The measurement method is validated in that the distribution of the time duration in which the signal attenuates by 5 dB is consistent with the existing statistical model and the range of the measured time duration in which the signal attenuation decreases from 5 dB to 0 dB is similar to that in the previous report.

In this paper, an asymptotic analysis of the inversely proportional setting (IPS) of carrier sense threshold (CST) and transmission power in densely deployed wireless local area networks (WLANs) is presented. In densely deployed WLANs, CST adjustment is a crucial technology to enhance spatial channel reuse, but it can starve surrounding transmitters due to an asymmetric carrier sensing relationship. In order for the carrier sensing relationship to be symmetric, the IPS of the CST and transmission power is a promising approach, i.e., each transmitter jointly adjusts the CST and transmission power in order for their product to be equal to those of others. By assuming that the set of potential transmitters follows a Poisson point process, the impact of the IPS on throughput is formulated based on stochastic geometry in two scenarios: an adjustment of a single transmitter and an identical adjustment of all transmitters. The asymptotic expression of the throughput in dense WLANs is derived and an explicit solution of the optimal CST is achieved as a function of the number of neighboring potential transmitters and signal-to-interference power ratio using approximations. This solution was confirmed through numerical results, where the explicit solution achieved throughput with a loss of less than 8% compared to the numerically evaluated optimal solution.

A potential game based distributed channel selection scheme is proposed in this paper to mitigate the flow-in-the-middle (FIM) throughput starvation problem that frequently occurs in dense wireless local area networks (WLANs). The FIM throughput starvation occurs when neighbors of a given node are not within the carrier sense ranges of each other. Since they spatially reuse the channel and at least one of them transmits with a high probability, the node in the middle would detect the channel being occupied for a prolonged time and therefore experience extremely low throughput. The basic idea of the proposed scheme is to let each access point (AP) select the channel that reduces the number of three-node chain topologies on its two-hop neighborhood contention graph. The proposed scheme is proved to be a potential game, i.e., the proposed scheme is guaranteed to converge. Graph-based simulation shows that starvation occurs on 20% of nodes when nodes randomly select their frequency channels. The proposed scheme significantly reduces the number of starved nodes along with iterations, outperforming the compared traditional potential game based scheme.

In database-driven spectrum sharing for 5G mobile networks, a primary user (PU) may experience harmful interference caused by unpredictable propagation paths, even when secondary users (SUs) follow a spectrum sharing policy established on the basis of a database. A framework for determining the optimal radius of a circular primary exclusive region (PER) on the basis of SU's information has been proposed. However, a practical PER can be complex-shaped and should be designed on the basis of the directivity of the PU antenna, and the SU information in each region. In this paper, we present a stochastic geometry analysis in a spatial grid-based spectrum database, and propose a design for an optimal complex-shaped PER. The database determines the transmission probability of the SUs on each divided annular sector. By regarding the SU's locations on each annular sector as an inhomogeneous Poisson point process, we analytically derive a PU's outage probability (OP), where the PU's OP is defined as the probability that the aggregate interference power at a PU from the SUs exceeds a threshold. Using the derived expression, we formulate an optimization problem to maximize the number of transmitting SUs, which optimizes the SU's transmission probability on each annular sector. Then, we numerically evaluate the solution of the optimization problem in various scenarios. The results show that the accuracy of the PER improves as the grid size decreases. In addition, we successfully design a complex-shaped PER with holes in which the SUs are permitted to transmit.

Secure communication is expected to be achieved by locating interfering transmitters near a transmitter sending confidential messages. This is because the transmitters in this arrangement would cause harmful interference to eavesdroppers. To investigate this proposal, this paper theoretically analyzes the secrecy outage probability (SOP) in clustered wireless networks. The SOP is formulated on the basis of the stochastic geometry approach, considering the case when the transmitters are distributed according to a Neyman-Scott cluster process. Numerical results reveal the impact of various clustered distributions of transmitters on the SOP. In addition, this paper conducts a comparative evaluation of the SOP between the clustered distribution and a uniform random distribution. It is revealed that the clustered distribution, in particular consisting of small size clusters and a large number of transmitters in each cluster, can reduce SOP compared to the uniform random distribution.

This paper presents a measurement of time-varying attenuation of IEEE802.11ad wireless LAN (WLAN) signals in 60GHz band induced by human blockage. The present measurement is a novel approach to obtain the attenuation, where a commercially available IEEE802.11ad access point (AP) and station (STA) are employed and the measurement is conducted under intermittent packet transmission. This paper also presents a hidden Markov model (HMM)-based signal power estimation scheme so that the attenuation is estimated from data obtained with a microwave spectrum analyzer which cannot detect signals of IEEE 802.11ad WLAN in itself. In this scheme, whether 11ad WLAN signals exist or not at each sampling instant is estimated based on HMM. Before the application of HMM, the scheme detects the number of HMM states via Bayesian information criterion and, thereby, prevents model over-fitting and consequent invalid power estimation. Our experiment revealed that the IEEE802.11ad WLAN signal attenuates by 5dB in a duration of 51.5ms when a human moves across the path between the AP and the STA at a velocity of 0.5m/s. This result is consistent with a previous report about an IEEE 802.11ad WLAN channel model.

The human blockage problem is an open issue in next- generation wireless access networks using millimeter-wave (mmWave) communications. A proactive base station (BS) handover system leveraging RGB and depth (RGB-D) cameras is proposed for solving the human blockage problem. RGB-D cameras observe mmWave communication ranges, and BS handover is conducted proactively before a human blockage causes serious performance degradation. However, this system must rely on a scheme that provides a guideline for selecting a BS to which the transfer can be done. In this paper, we propose a mmWave throughput estimation scheme using an online machine learning algorithm and depth images obtained by the RGB-D camera. The algorithm learns the relationship between depth images and measured throughputs, and estimates throughput from depth images. The scheme enables the handover system to estimate throughput quickly and adaptively to the wireless environment without transmitting any control frames. We conducted a proof-of-concept experiment by using a testbed consisting of IEEE 802.11ad mmWave wireless local area network devices and an RGB-D camera. The experiment confirmed that the proposed scheme estimates throughput from a depth image with an RMS error of 114-178 Mbit/s in real time.

In database-driven spectrum sharing, despite the spectrum sharing policy given by a database, harmful interference can occur between a primary user (PU) and a secondary user (SU) due to the unexpected propagation paths. In a previous study, a primary exclusive region (PER) centered at a PU, wherein the SUs are forbidden to use the spectrum, has been proposed. However, the PER figure that efficiently covers the regions where interference occurs, cannot be circular. In this paper, we propose a framework for updating the PER adaptively with machine learning, when interference occurs. The framework employs undersampling and oversampling schemes considering the propagation characteristics and shadow fading in order to solve an imbalanced data problem degrading estimation accuracy of appropriate shape of PER. Our simulation results demonstrate that the area of PER with the proposed framework is smaller by 54% than that of the fixed circular PER setting, and the proposed sampling scheme achieves a 1% interference probability with 21% fewer iterations and a 6% smaller area compared to the existing sampling schemes.

In database-driven spectrum sharing for 5G mobile networks, a primary user (PU) can experience harmful interference due to unexpected propagation paths, even when the secondary user (SU) follows the spectrum sharing policy stored in a database, during operations. A framework for deriving the optimal radius of a circular primary exclusive region (PER) has been proposed. However, a practical PER can be non-circular and should be designed on the basis of the geographical information and directivity of the PU antenna. To design a non-circular PER, in this paper, we introduce a spatial grid-based spectrum database; the database permits or forbids SU transmissions for each annular sector divided by a spatial grid. Considering the random locations of the SUs on each annular sector as an inhomogeneous Poisson point process, we analytically derive the PU's outage probability (OP) using stochastic geometry, where the PU's OP is defined as the probability that the aggregate interference power at a PU, from the SUs, exceeds a threshold. Numerical results show that a non- circular PER results in a lower PU OP compared to a circular PER.

This paper proposes a decentralized scheme to improve the energy efficiency (EE) in dense wireless local area networks (WLANs), based on inversely proportional setting (IPS) of carrier sense threshold (CST) and transmit power. Jointly adjusting the CST and transmit power in an inversely proportional manner, i.e., keeping their product to be a constant, is a promising technique to increase transmit opportunities in dense WLANs, without generating unfairness issue of asymmetric carrier sensing. The proposed scheme is formulated in the non-cooperative game framework. Two utility functions are designed, where the first utility function is an approximation of the EE. To guarantee the existence of the equilibrium, and to ensure the convergence of the proposed scheme, we further incorporate potential games, based on which the second utility function is de- signed. The second utility function can be considered as assigning a cost on the first utility function. Simulation results show that the proposed scheme provides considerable improvement in EE and average throughput under various node densities. The second utility function outperforms the first utility function, while the first one requires only local information to be calculated.

This paper provides an appropriate allocation scheme of channel resources for drone communications. In recent years, the usage of drones has been increasing for a wide range of applications, and their communications require using additional frequency bands. However, their communications must avoid interference with primary users (e.g., radar systems) when they use additional frequency bands. We assume that their communications use two frequency bands, the main band and the backup band, to avoid interference. The proposed resource allocation method, which determines whether each drone should use the main or the backup bands, enables drone communications to use the main band efficiently without causing interference. For the proposed resource allocation scheme, this paper presents a stochastic geometry analysis of interference in drone networks. Based on the assumption that the distribution of drones follows a 3D Poisson point process, we analytically derive the radar's outage probability (OP) and the drone's OP. From these analyses, an optimization problem is formulated to maximize the number of drones using the main band, since drones can transmit massive amounts of data by using the main band. Then, we numerically evaluate the solution of the optimization problem. Our results show that the maximum ratio of the number of drones using the main band to the total number of all drones increases along with the size of the primary exclusive region.

Inversely proportional setting (IPS) of the transmission power and carrier sense threshold (CST) to a portion of access points/stations is discussed through numerical evaluation and experiments to examine coexistence with legacy devices. In densely deployed wireless local area networks (WLANs), tuning CST is a promising approach to facilitate spatial channel reuse. Particularly, IPS of the transmission power and CST keeps symmetric carrier sensing relationship between any two transceivers, and thus provides a novel solution for the starvation due to asymmetric carrier sensing relationship. We first model the throughput of two transmitter-receiver pairs when applying IPS. In addition, through experiments, we verify the throughput model and confirm that IPS does not cause throughput starvation even when applying IPS only to some APs and/or STAs.

In millimeter wave (mmWave) communications, when a pedestrian blocks the path of the line of sight (LOS) between a station (STA) and a base station (BS), the quality of communication sharply deteriorates. In order to cope with such human blockage, an RGB and depth (RGB-D) camera-assisted proactive handover scheme is proposed in this paper. The scheme uses images from an RGB-D camera to predict human blockage, and allows the selection of an appropriate BS based on this prediction. To clarify the impact of the accuracy of blockage prediction on throughput performance and assess the ideal performance of the proposed scheme, we propose the performance modeling of both proactive and reactive handover schemes based on the received power level. Numerical evaluation results revealed conditions under which the proactive handover scheme yields higher spectral efficiency than the reactive scheme. We conducted simulations to verify the performance gain of the proposed scheme in a realistic scenario where the LOS paths between a mobile STA and the BS were stochastically blocked. The results show that the proactive handover scheme can reduce the duration of outages due to human blockage, and increased system throughput by 12.1% over the reactive handover scheme.

To realize high-speed and high-capacity 5G mobile networks, secondary users (SUs) may opportunistically use licensed spectrum allocated to primary users (PUs). This paper presents a framework to deal with the situation that occurs when a PU suffers harmful interference caused by the secondary use of the spectrum allocated to it. Here, the PU is assumed to be a radar system. In our proposed framework, the PU informs the database that it is suffering harmful interference. Receiving this information, the database updates the primary exclusive region (PER), where SUs are prohibited from using the licensed spectrum. The updated PER depends on the current knowledge of the SUs, which is stored in the database. We assume a circular PER centered at the primary receiver (PR) and derive its optimal radius using stochastic geometry. For each type of SU knowledge stored in the database, we evaluate the optimal PER radius for a target probability at which the PU suffers harmful interference. The results show that more detailed knowledge of the SU density and transmission powers leads to a smaller updated PER radius. Hence, a more efficient spatial reuse of the licensed spectrum is achieved.

This paper proposes frame length optimization for wireless local area networks (WLANs) using an in-band full-duplex system that enables a WLAN access point and stations to transmit and receive frames at the same time on the same frequency channel. In in-band full-duplex WLANs, a primary sender which captures the channel transmits a frame to the intended receiver called a secondary sender and then the secondary sender transmits a frame reacting to the primary sender's transmission. The difference of time length of frames transmitted by the primary sender and the secondary sender wastes the frequency channel where more frames could be transmitted. The wasted time decreases the system throughput performance of the in-band full-duplex system. In order to solve this problem, we propose a scheme where the secondary sender adjusts the length of its frame to the length of the primary sender's frame by selecting frames used for frame aggregation properly. We evaluate the average delay, the average wasted time and the system throughput performance by computer simulations. The simulation results show that the proposed optimization reduces the delay by 49%, reduces the wasted time by 99.9% and improves the system throughput performance by 15% when the traffic is saturated.

Interference management is indispensable for dense wireless networks to improve system performance. Reduction in spatial channel overlap is one of the most effective approaches, which leads to an increase in the transmission opportunity of each access point (AP) and mitigation of channel interference at a reception point. In this paper, coverage overlap is modeled as a subset of two-dimensional Euclidean plane, and a joint decentralized scheme of transmission power control (TPC) and dynamic channel assignment (DCA) is described for reduction in the coverage overlap. In particular, the DCA scheme can be formulated as a potential game in which unilateral improvement dynamics are guaranteed to converge to a Nash equilibrium. The novel feature of this paper is that spatial adaptive play (SAP) is employed for channel update algorithm to achieve more improvement of the reduction in the total channel overlap area. Simulation results show that the scheme based on SAP is more effective to reduce the total channel overlap area than best response based learning algorithm.

This paper presents implementation of a reactive base station selection scheme for millimeter-wave (mmWave) communications. In mmWave communications, the frame loss rate increases and the throughput sharply decreases when a pedestrian blocks a line-of-sight (LOS) path. To alleviate this human blockage problem, base stations can be selected so as to maintain LOS paths on the basis of communication quality. In this paper, we build a testbed using off the shelf IEEE 802.11ad based wireless local area network (WLAN) devices, and implement a reactive base station selection scheme on the testbed. To the best of our knowledge, there is no existing work which experimentally evaluates human blockage detection system using actual IEEE 802.11ad devices. Our prototype system monitors the throughput measured at each base station and detects human blockage when the throughput decreases below a threshold. The human blockage detection triggers the base station switching. Our experimental results show that the reactive base station selection scheme decreases the duration in which human blockage degrades throughput performance, and the total amount of received data increases by 21%.

In-band full-duplex (IBFD) operation can potentially double the spectral efficiency of wireless networks. For IBFD operation, self-interference is a critical issue. In addition, in full-duplex cellular (FDC) networks, particularly when the cell size is small, inter-user interference would be another limiting factor for the performance. To overcome these issues, the scheduling scheme proposed in this paper is to adaptively utilize bidirectional IBFD in addition to half-duplex (HD) and unidirectional IBFD in FDC networks according to the residual self-interference after interference cancellation and inter-user interference. The proposed scheme is based on generalized proportional fair scheduling by using a fairness parameter. Extensive simulations are conducted to analyze the impact of the cell size. Simulation results revealed that the use of bidirectional IBFD is extremely effective for a small cell with few users if self-interference is sufficiently canceled because inter-user interference is large in small cell and the scheduler tends to select bidirectional IBFD in the cell with few users.

Wireless local area network (WLAN) using millimeter wave (mmWave) communications is one of enabler of next generation wireless access networks since huge bandwidth is available in mmWave bands and it enables beyond Gbit/s communications. This paper experimentally evaluate the PHY rate and the throughput performance of off the shelf mmWave WLAN devices that are compatible with IEEE 802.11ad standard. Our experimental evaluation shows that the association can be established until the distance between an access point (AP) and a station (STA) is 27 m and even if a STA is located at opposite side of an AP, the association still be established until the distance is smaller than 5 m. Our interference evaluation reveals an unfairness of throughput between two pairs of the AP and the STA.

This paper proposes an RGBD cameras-based wireless environment prediction and presents an implementation of a proactive traffic control system using the prediction results. In millimeter-wave and terahertz communications, frame loss rate increases and throughput is decreased tremendously when pedestrians block line-of-sight paths. Such serious human blockage does not occur in microwave communications. By using the RGBD cameras, we can predict human blockage easily as the cameras can detect obstacles and their mobility. We implement a prototype of a proactive traffic control system using camera-based human blockage prediction to demonstrate the feasibility of the proposed system.

This paper proposes a proactive base station selection system for millimeter-wave (mmWave) communications based on human blockage prediction using RGB and depth (RGB-D) cameras. In mmWave communications, the frame loss rate increases and the throughput sharply decreases if a pedestrian blocks a line-of-sight (LOS) path between a station and a base station.To address this human blockage problem, multiple base stations can be arranged so as to maintain at least one LOS path.For base station selection in particular, RGB-D camera images can be used to estimate the mobility of pedestrians and to predict when blockage of LOS paths will occur. Using IEEE 802.11ad-based wireless local area network (WLAN) devices, a testbed for implementing the proposed system was built. The results of experiments on the influence of human blockage confirmed the presence of significant throughput degradation due to human blockage.Furthermore, the innovative experimental results demonstrated that the proactive base station selection system can considerably reduce the duration of human blockage-induced degradation of throughput performance relative to reactive base station selection systems based on throughput performance.

Dense wireless networks require advanced interference management for the performance improvement. In carrier sense multiple access with collision avoidance (CSMA/CA) networks, the mutual interference can be modeled as the overlap area, and range adjustment, such as transmission power control (TPC) and dynamic channel assignment (DCA), is effective to reduce the overlap area. A game-theoretic framework is utilized to construct a joint TPC and DCA scheme which reduces the total overlap area. Although the evaluation of the overlap area requires topology of neighboring APs, i.e., received power levels among neighboring APs are required, the proposed scheme does not require such information because decomposed overlap areas between any two APs are utilized. Moreover, unilateral improvement dynamics of channel assignment are guaranteed to converge to a Nash equilibrium because the game is shown to be a potential game. We evaluate the proposed scheme through simulations and experiments. Simulation results reveal that the proposed joint TPC and DCA scheme effectively reduces the total overlap area compared to another game-theoretic interference management scheme particularly when the number of APs is large. It is because the proposed scheme can efficiently manage the coverage overlap even when there is a variation in transmission power level. Experimental results demonstrate the convergence of the proposed scheme in a real environment.

In dense wireless local area network (WLAN) environments, the haphazard deployment of WLAN devices causes them to overlap. In such situations, a coverage adaptation through transmission power control (TPC) is required. At the same time, coverage holes should be taken into consideration for connectivity maintenance among devices. In this paper, a TPC scheme is proposed that reduces overlaps in an autonomous manner, without generating target coverage holes. The problem is formulated as a target coverage problem, in which wireless stations (STAs) are treated as target points, and the transmission power is appropriately controlled by the proposed scheme. In addition, we prove that the target coverage game is a potential game, and the proposed method is guaranteed to converge to a Nash equilibrium after a finite number of iterations. In order to prevent the occurrence of coverage holes that cannot be detected by the available STA information, we propose the exclusion of coverage boundary access points (APs) from reducing transmission power, with the idea that each boundary AP always contains the region covered by it alone. The coverage boundary APs are determined through Delaunay triangulation of all APs. The proposed scheme is evaluated by performing some simulations. The simulation results indicate that the proposed TPC scheme is effective in reducing coverage overlaps without generating coverage holes.

An adaptive channel allocation scheme for bidirectional in-band full-duplex networks is proposed by modifying a channel allocation scheme for half-duplex networks on the basis of the property of bilateral symmetric interaction. To adapt schemes for half-duplex networks to full-duplex networks, we modify them so that channel update process successfully converges to a Nash equilibrium, i.e., it does not cycle. Specifically, the sum of products of transmission power and received interference power (instead of the sum of received interference power) at a communication pair is used as the novel payoff function. Then, the proposed channel allocation scheme is proved to be a potential game, and thus, the game is guaranteed to converge in a finite number of steps. Simulation results reveal that the proposed scheme is able to effectively allocate channels in bi-directional in-band full-duplex networks.

Inserting attenuators between transceivers and antennas is proposed to improve spatial channel reuse in carrier sense multiple access with collision avoidance based WLANs. Using attenuators enables the access points or stations to decrease the powers of transmission signals and received signals, which equivalently results in the increase of the carrier sense threshold. Thus, the attenuation value control enables inversely proportional setting of transmission power and carrier sense threshold, which is known to provide a novel solution for the unfairness problem caused by variable transmission power or variable carrier sense threshold. We first derive a simple and useful sufficient condition such that the aggregate spectral efficiency is higher when attenuators are used than they are not by using some approximations. Through a numerical evaluation as well as a testbed for using attenuators, the sufficient condition is shown to be valid despite the approximations and throughput improvement is verified. The result from the testbed also discloses a new unfairness problem due to an individual difference in transmission power or carrier sense threshold.

Bandwidth barter is an effective way of satisfying throughput requirements in wireless networks; we could expect a station (STA) allows another STA to borrow its bandwidth as long as it is also beneficial for the STA. Our previous work proved bandwidth barter between two STAs is optimized based on Nash bargaining solution (NBS), which brings the Pareto efficiency and the proportional fairness in the bartering game. However, it still remains an open issue how to solve the bartering game when the number of STAs is N (N >; 2), which is discussed in this paper.

We propose a TCP window-size delegation method for TXOP Exchange applicable to the downlink in wireless access networks. In TXOP Exchange, the compliant stations (STAs) cooperatively use their available bandwidth in accordance with their required QoSs. TXOP Exchange was previously validated for the uplink. The proposed delegation method enables STAs to delegate their bandwidth for the downlink as well without requiring any modifications to the legacy access point or the STAs. Simulation demonstrated that this method works well.