Dynamic Graph Convolutional Neural Networks (DGCNNs) are effective for processing graph-structured data, particularly point clouds used in immersive media, augmented reality (AR), and virtual reality (VR) applications. However, their inference can be computationally intensive, especially with high-resolution inputs, posing challenges for latency-sensitive and resource-constrained deployments. To address these challenges, we propose confidence-based early exit architectures tailored for fast inference in multimedia-oriented DGCNN systems. The proposed method accelerates inference by allowing samples to exit at intermediate layers once a confidence threshold is met, while preserving or even enhancing classification accuracy. Our best-performing configuration achieves a 3× speed-up alongside a 0.59% accuracy gain, while the most efficient design provides a 4.3× speed-up with only a minor accuracy trade-off. These results demonstrate the potential of early exiting as a lightweight and scalable solution for point-cloud inference in future multimedia communication systems.

Foundation deep learning (DL) models are general models, designed to learn general, robust and adaptable representations of their target modality, enabling finetuning across a range of downstream tasks. These models are pretrained on large, unlabeled datasets using self-supervised learning (SSL). Foundation models have demonstrated better generalization than traditional supervised approaches, a critical requirement for wireless communications where the dynamic environment demands model adaptability. In this work, we propose and demonstrate the effectiveness of a Vision Transformer (ViT) as a radio foundation model for spectrogram learning. We introduce a Masked Spectrogram Modeling (MSM) approach to pretrain the ViT in a selfsupervised fashion. We evaluate the ViT-based foundation model on two downstream tasks: Human Activity sensing and Spectrogram Segmentation. Experimental results demonstrate competitive performance to supervised training while generalizing across diverse domains. Notably, the pretrained ViT model outperforms a four-times larger model that is trained from scratch on the spectrogram segmentation task, while requiring significantly less training time, and achieves competitive performance on the human activity sensing task. This work demonstrates the effectiveness of ViT with MSM for pretraining as a promising technique for scalable foundation model development in future 6G networks.

The deployment of deep learning (DL) models on Internet of Things (IoT) devices is constrained by limited computational resources, necessitating effective model compression techniques. This paper introduces a novel, structured, nonuniform, block-wise compression approach, tailored for resourceconstrained IoT applications that require Angle of Arrival (AoA) estimation. Our method prunes all model blocks of layers, adjusting the pruning intensity based on each block's measured contribution to overall performance: less critical blocks undergo more aggressive pruning, while essential blocks are preserved to maintain accuracy. Applied to MobileNetV3 for AoA estimation, our technique achieved a 35.9X reduction in parameter count, an 11.5X reduction in model size, a 7.7X reduction in multiplyaccumulate operations (MACs), a 4.7X improvement in inference speed on CPU, and a 2.4X improvement on GPU, with minimal impact on accuracy. These results are achieved on a real-world dataset collected on a software defined radio (SDR) testbed to validate the effectiveness of the proposed solution. This demonstrates that our approach can achieve a highly favorable trade-off between compression and performance, supporting the deployment of tiny AoA models in latency-critical and powersensitive IoT environments.

Artificial intelligence (AI) is anticipated to play a pivotal role in 6G. However, a key challenge in developing AIpowered solutions is the extensive data collection and labeling efforts required to train supervised deep learning models. To overcome this, self-supervised learning (SSL) approaches have recently demonstrated remarkable success across various domains by leveraging large volumes of unlabeled data to achieve nearsupervised performance. In this paper, we propose an effective SSL scheme for radio signal representation learning using momentum contrast. By applying contrastive learning, our method extracts robust, transferable representations from a large realworld dataset. We assess the generalizability of these learned representations across two wireless communications tasks: angle of arrival (AOA) estimation and automatic modulation classification (AMC). Our results show that carefully designed augmentations and diverse data enable contrastive learning to produce highquality, invariant latent representations. These representations are effective even with frozen encoder weights, and fine-tuning further enhances performance, surpassing supervised baselines. To the best of our knowledge, this is the first work to propose and demonstrate the effectiveness of self-supervised learning for radio signals across multiple tasks. Our findings highlight the potential of self-supervised learning to transform AI for wireless communications by reducing dependence on labeled data and improving model generalization - paving the way for scalable foundational 6G AI models and solutions.

The Internet of Things (IoT) integrates deep learning (DL) to enhance real-time data processing across diverse applications. However, deploying DL models on resourceconstrained IoT devices remains challenging, especially for tasks such as Angle-of-Arrival (AoA) estimation in dynamic environments. Variations in deployment conditions, such as changing modulation schemes, lead to domain shifts that degrade traditional models' performance, underscoring the need for adaptive, low-complexity DL frameworks. This paper introduces a novel compact phase and amplitude representation within a Prototypical Network-based approach, optimized for domain-adaptive AoA prediction in IoT and validated using real data from a softwaredefined radio (SDR) testbed. Compared to covariance and raw IQ data, our proposed representation reduces Mean Absolute Error (MAE) by approximately 32 % and 55 %, respectively, in unseen modulation scenarios. Further, evaluations on an SDR dataset collected using a$2 \times 2$Uniform Rectangular Array (URA) configuration with seven modulation schemes demonstrate that Prototypical Networks with few-shot learning enable accurate and robust adaptation with minimal data, maintaining high accuracy across both seen and unseen modulations

Foundation Models (FMs) have surged in popularity over the past five years, with applications spanning fields from computer vision to natural language processing. At the same time, Brain-Computer Interfaces (BCIs) have also gained momentum due to their potential to support individuals with complex disabilities. Among various BCI paradigms, code-modulated Visual Evoked Potentials (c-VEPs) remain relatively understudied, despite offering high information transfer rates and large selection target capacities. However, c-VEP systems require lengthy calibration sessions, significantly limiting their practicality, particularly outside of laboratory settings. In this study, we use a FM for the first time to eliminate the need for lengthy calibration in c-VEP BCI systems. We evaluated two approaches: (1) a truly calibration-free approach requiring no subject-specific data, and (2) a limited calibration approach, where we assessed the benefit of incorporating incremental amounts of calibration data. In both cases, a classification head is trained on data from other subjects. For a new subject, no calibration data is required in the calibration-free setup, making the c-VEP system effectively plug-and-play. The proposed method was tested on two c-VEP datasets. For the calibration-free approach, the average accuracy on the first dataset (n = 17) was 68.8% ± 17.6%, comparable to the full-calibration performance reported in the original study (66.2% ± 13.8%), which required approximately 11 minutes of calibration. On the second dataset (n = 12), the calibration-free accuracy was 71.8% ± 20.2%, versus 93.7% ± 5.5% from the original study, which required around 3.5 minutes. A limited-calibration approach using only 20% of the subject’s data (approximately 43 seconds) yielded 92%±5.2% accuracy. These results indicate that our FM–based approach can effectively eliminate or significantly reduce the need for lengthy calibration in c-VEP BCIs.

Automatic Modulation Classification (AMC) is essential for wireless communication systems, particularly in resource-constrained environments such as vehicular networks. However, deploying Deep Neural Networks (DNNs) on receivers is challenging due to computational and energy constraints. To address this, we propose the first joint integration of early exit and structured pruning for AMC, optimizing computational efficiency without significantly compromising accuracy. Through theoretical analysis, we demonstrate the unique interactions between these techniques and the non-intuitive impact of pruning on early-exit models. Our experiments evaluate the impact of various pruning criteria, pruning percentages, and early-exit thresholds. Our results reveal that the effectiveness of different pruning criteria in early-exit AMC models is highly context-dependent, with no universally optimal strategy, and that increasing pruning levels and early-exit thresholds intensifies the trade-off between computational efficiency and classification accuracy.

Despite the heightened popularity of Virtual Re-ality (VR), streaming high-resolution stereoscopic VR remains a challenge. This is primarily due to significant bandwidth demands of high-definition VR content. While advanced deep neural networks (DNN s) have demonstrated the potential to outperform standard codecs, their integration into real-world transmission frameworks is complex and not directly compatible with current encoding standards. To bridge this gap, this paper proposes a novel technique for compressing and transmitting 8K stereoscopic scenes using Variational Autoencoder (VAE) GAN latents represented as 3-channel RGB scenes that can be transmitted via standard encoders. The proposed method reduces bandwidth requirements by 45.1 % across different 8K scenes while maintaining visual quality, highlighting the effectiveness of the approach. This study also investigates the impact of varying patch-sizes of input frames for model training and evaluate its influence on client-side reconstructions. We then explore various transmission configurations of latent frames. Our findings suggest that while residual transmission offers limited benefits for 3-channel latent frame compression, raw transmission consistently yields better results, particularly for texture-heavy scenes. To the best of our knowledge, this is the first such transmission study on 8K stereoscopic scenes for cloud-based VR, providing valuable insights for optimizing high-resolution VR streaming systems. Code: github.com/sampreetucalgary07/8K-VR-compression.

Deploying large-scale foundation models (FMs) in resource-constrained devices presents critical challenges due to their substantial computational and memory requirements. This is particularly relevant for multi-task wireless sensing FMs running on sensors. To overcome these limitations, we propose a tiny federated wireless foundation model (WFM) framework that combines spectrogram-guided structured block-wise pruning with federated learning (FL) for efficient on-device deployment. Our approach prunes non-essential encoder blocks in vision transformers (ViTs) by leveraging the masked spectrogram modeling (MSM) pretraining loss as an importance indicator, ensuring only the most structurally significant components are retained. This enables federated adaptation with frozen backbones and lightweight, task-specific heads, minimizing both computational burden and communication overhead. The pruning strategy preserves the integrity of spectrogram reconstruction, while federated fine-tuning supports decentralized learning across clients with heterogeneous data distributions. Experimental results on human activity sensing and radio signal identification tasks confirm the efficacy of our approach. Specifically, the pruned ViT-based WFMs achieve up to 93% multiply-accumulate operations (MACs) reduction, 85% lower CPU inference time, and 49% reduction in communication overhead, all while maintaining high task accuracy. Our method demonstrates strong generalization and robustness across varying pruning ratios and data heterogeneity levels, while substantially reducing communication overhead, making it highly suitable for real-world industrial IoT deployments.

Foundational deep learning (DL) models are general models, trained on large, diverse, and unlabelled datasets, typically using self-supervised learning techniques - and have led to significant advancements especially in natural language processing. These pretrained models can be fine-tuned for related downstream tasks, offering faster development and reduced training costs, while often achieving improved performance. In this work, we introduce Masked Spectrogram Modeling, a novel self-supervised learning approach for pretraining foundational DL models on radio signals. Adopting a Convolutional LSTM architecture for efficient spatio-temporal processing, we pretrain the model with an unlabelled radio dataset collected from over-the-air measurements. Subsequently, the pretrained model is fine-tuned for two downstream tasks: spectrum forecasting and segmentation. Experimental results demonstrate that our methodology achieves competitive performance in both forecasting accuracy and segmentation, validating its effectiveness for developing foundational radio models.

Deep learning (DL) techniques have recently emerged to efficiently manage mmWave beam transmissions without requiring time consuming beam sweeping strategies. A fundamental challenge in these methods is their dependency on hardware-specific training data and their limited ability to generalize. Large drops in performance are reported in literature when DL models trained in one antenna environment are applied in another. This paper proposes the application of Prototypical Networks to address this challenge – and utilizes the DeepBeam real-world dataset [1] to validate the developed solutions. Prototypical Networks (PN) excel in extracting features to establish class-specific prototypes during the training, resulting in precise embeddings that encapsulate the defining features of the data. We demonstrate the effectiveness of PN to enable generalization of deep beam predictors across unseen antennas. Our approach, which integrates data normalization and prototype normalization with the PN, achieves an average beam classification accuracy of 74.11% when trained and tested on different antenna datasets. This is an improvement of 398% compared to baseline performances reported in literature that do not account for such domain shifts. To the best of our knowledge, this work represents the first demonstration of the value of Prototypical Networks for domain adaptation in wireless networks, providing a foundation for future research in this area.

Over-the-air federated learning (OTA-FL) is a distributed machine learning technique where multiple devices collaboratively train a shared model without sharing their raw data with a central server. The devices exchange model updates concurrently over-the-air and they are aggregated without the need for dedicated wireless resources for each device. A major challenge in OTA-FL is that edge devices are limited in their computation, energy, and communication resources. To address this, we investigate how deep neural network compression techniques can be applied in an OTA-FL system. We propose a compression pipeline comprised of pruning and quantization-aware training that significantly reduces both the computation and communication requirements while maintaining an on-par accuracy to the uncompressed models. We thoroughly investigate the reduction in model size, accuracy, and convergence when pruning and quantization-aware training are applied individually and collectively at multiple pruning and quantization levels. Detailed experiments are conducted on two-different deep learning models and two-different datasets, and under varying signal-to-noise ratios (SNRs) and numbers of clients. We then thoroughly present and discuss the resulting trade-offs and findings. Our results demonstrate that deep compression is very effective in an OTA-FL system and negligible losses in accuracy are possible while maintaining up to 80 percent reductions in model size.

The open radio access network (O-RAN) architecture supports intelligent network control algorithms as one of its core capabilities. Data-driven applications incorporate such algorithms to optimize radio access network (RAN) functions via RAN intelligent controllers (RICs). Deep reinforcement learning (DRL) algorithms are among the main approaches adopted in the O-RAN literature to solve dynamic radio resource management problems. However, despite the benefits introduced by the O-RAN RICs, the practical adoption of DRL algorithms in real network deployments falls behind. This is primarily due to the slow convergence and unstable performance exhibited by DRL agents upon deployment and when encountering previously unseen network conditions. In this paper, we address these challenges by proposing transfer learning (TL) as a core component of the training and deployment workflows for the DRL-based closed-loop control of O-RAN functionalities. To this end, we propose and design a hybrid TL-aided approach that leverages the advantages of both policy reuse and distillation TL methods to provide safe and accelerated convergence in DRL-based O-RAN slicing. We conduct a thorough experiment that accommodates multiple services, including real VR gaming traffic to reflect practical scenarios of O-RAN slicing. We also propose and implement policy reuse and distillation-aided DRL and non-TL-aided DRL as three separate baselines. The proposed hybrid approach shows at least: 7.7% and 20.7% improvements in the average initial reward value and the percentage of converged scenarios, and a 64.6% decrease in reward variance while maintaining fast convergence and enhancing the generalizability compared with the baselines.

We consider federated learning in a wireless edge network, where multiple power-limited mobile devices collaboratively train a global model, using their local data with the assistance of an edge server. Exploiting over-the-air computation, the edge server updates the global model via analog aggregation of the local models over noisy wireless fading channels. Unlike existing works that separately optimize computation and communication at each step of the learning algorithm, in this work, we jointly optimize the training of the global model and the analog aggregation of the local models over time. Our objective is to minimize the accumulated training loss at the edge server, subject to individual long-term transmit power constraints at the mobile devices. We propose an efficient algorithm, termed Online Model Updating with Analog Aggregation (OMUAA), to adaptively update the local and global models based on the time-varying communication environment. The trained model of OMUAA is channel-and power-aware, and it is in closed form incurring low computational complexity. We study the mutual impact between model training and analog aggregation over time, to derive performance bounds on the computation and communication performance metrics. Furthermore, we consider a variant of OMUAA with double regularization on both the local and global models, termed OMUAA-DR, and show that it can significantly reduce the convergence time to reach long-term transmit power constraints. In addition, we extend both OMUAA and OMUAA-DR to enable analog gradient aggregation, while preserving their performance bounds. Simulation results based on real-world image classification datasets and typical wireless network settings demonstrate substantial performance gain of OMUAA and OMUAA-DR over the known best alternatives.

We consider cooperative multiple-input multiple-output (MIMO) precoding design with multiple access points (APs) assisted by a central controller (CC) in a fading environment. Even though each AP may have its own local channel state information (CSI), due to the communication delay in the backhaul, neither the APs nor the CC has timely global CSI. Under this hierarchical semi-online setting, our goal is to minimize the accumulated precoding deviation, between the actual local precoders executed by the APs and an ideal cooperative precoder based on timely and perfect global CSI, subject to per-AP transmit power limits. We propose an efficient algorithm, termed Semi-Online Precoding with Information Parsing (SOPIP), which accounts for the network heterogeneity in information timeliness and computational capacity. SOPIP does not require the CC to send the full global CSI to each AP. Instead, it takes advantage of the precoder structure to substantially lower the communication overhead, while allowing each AP to effectively combine its own timely local CSI with the delayed global CSI to enable adaptive precoder updates. We analyze the performance of SOPIP in the presence of multi-slot communication delay, CSI inaccuracy, and gradient estimation error, showing that it has a bounded performance gap from an offline optimal solution. Simulation results under typical cellular system settings further demonstrate the substantial performance gain of SOPIP over other centralized and distributed schemes.

Automatic modulation classification (AMC) plays a critical role in wireless communications by autonomously classifying signals transmitted over the radio spectrum. Deep learning (DL) techniques are increasingly being used for AMC due to their ability to extract complex wireless signal features. However, DL models are computationally intensive and incur high inference latencies. This paper proposes the application of early exiting (EE) techniques for DL models used for AMC to accelerate inference. We present and analyze four early exiting architectures and a customized multi-branch training algorithm for this problem. Through extensive experimentation, we show that signals with moderate to high signal-to-noise ratios (SNRs) are easier to classify, do not require deep architectures, and can therefore leverage the proposed EE architectures. Our experimental results demonstrate that EE techniques can significantly reduce the inference speed of deep neural networks without sacrificing classification accuracy. We also thoroughly study the trade-off between classification accuracy and inference time when using these architectures. To the best of our knowledge, this work represents the first attempt to apply early exiting methods to AMC, providing a foundation for future research in this area.

The success of immersive applications such as virtual reality (VR) gaming and metaverse services depends on low latency and reliable connectivity. To provide seamless user experiences, the open radio access network (O-RAN) architecture and 6G networks are expected to play a crucial role. RAN slicing, a critical component of the O-RAN paradigm, enables network resources to be allocated based on the needs of immersive services, creating multiple virtual networks on a single physical infrastructure. In the O-RAN literature, deep reinforcement learning (DRL) algorithms are commonly used to optimize resource allocation. However, the practical adoption of DRL in live deployments has been sluggish. This is primarily due to the slow convergence and performance instabilities suffered by the DRL agents both upon initial deployment and when there are significant changes in network conditions. In this paper, we investigate the impact of time series forecasting of traffic demands on the convergence of the DRL-based slicing agents. For that, we conduct an exhaustive experiment that supports multiple services including real VR gaming traffic. We then propose a novel forecasting-aided DRL approach and its respective O-RAN practical deployment workflow to enhance DRL convergence. Our approach shows up to 22.8%, 86.3%, and 300% improvements in the average initial reward value, convergence rate, and number of converged scenarios respectively, enhancing the generalizability of the DRL agents compared with the implemented baselines. The results also indicate that our approach is robust against forecasting errors and that forecasting models do not have to be ideal.

Cloud-based Virtual Reality (VR) gaming is gaining popularity to provide immersive experiences without requiring bulky hardware. However, managing network resources for these games is crucial to prevent subpar user experience and unnecessary costs. Predicting VR traffic patterns, such as video frame sizes can enable proactive network resource allocation and lead to improved quality of service (QoS). To this end, in this paper we first evaluate the efficacy of various Machine Learning (ML) models for predicting gaming traffic frame sizes using data collected from a real-world cloud-based VR game testbed. We then investigate the effectiveness of transfer learning (TL) in predicting frame size traffic patterns across different games and network conditions using an online learning method that we propose. The findings show that using the TL approach for online learning prediction can reduce overall traffic prediction error by up to 54%. Overall, this paper contributes to the understanding of cloud-based VR traffic patterns and can be of interest to developers, practitioners, and researchers interested in optimizing the performance of such systems.

The recent advances in machine learning and deep neural networks have made them attractive candidates for wireless communications functions such as channel estimation, decoding, and downlink channel state information (CSI) compression. However, most of these neural networks are large and inefficient making it a barrier for deployment in practical wireless systems that require low-latency and low memory footprints for individual network functions. To mitigate these limitations, we propose accelerated and compressed efficient neural networks for massive MIMO CSI feedback. Specifically, we have thoroughly investigated the adoption of network pruning, post-training dynamic range quantization, and weight clustering to optimize CSI feedback compression for massive MIMO systems. Furthermore, we have deployed the proposed model compression techniques on commodity hardware and demonstrated that in order to achieve inference gains, specialized libraries that accelerate computations for sparse neural networks are required. Our findings indicate that there is remarkable value in applying these model compression techniques and the proposed joint pruning and quantization approach reduced model size by 86.5% and inference time by 76.2% with minimal impact to model accuracy. These compression methods are crucial to pave the way for practical adoption and deployments of deep learning-based techniques in commercial wireless systems.

Predicting packet volume accurately plays a vital role in network management and optimization. This study proposes a novel approach combining a hybrid wavelet transform and a deep stacking ensemble model to forecast packet volume. The proposed model leverages the wavelet transform's strengths and deep stacking ensemble techniques to enhance the predictive performance. The univariate prediction is conducted for different horizons, including 1, 3, 7, and 10 steps ahead. The wavelet transform is employed to capture the time-frequency characteristics of the packet volume data, enabling a more comprehensive analysis. The model can extract local and global features by decomposing the time series into different scales using wavelet analysis, improving prediction accuracy. The deep stacking ensemble model is utilized to leverage the collective intelligence of multiple base models. Through a series of stacking layers, the model learns to combine the predictions of individual models, allowing for a more robust and accurate forecast. The proposed hybrid wavelet transform and deep stacking ensemble model was subjected to experimental evaluations using real-world packet volume datasets. The outcomes of these evaluations exhibit the effectiveness of the model. The model achieves superior predictive performance compared to traditional methods, showcasing its potential for practical network management and optimization applications.

Children are an under-served population in the field of brain-computer interface (BCI) development. The high prevalence of lifelong disability coupled with the diversity and plasticity of children's brains make them ideal candidates for personalized BCI systems. Channel selection methods provide a tool for the in-session personalization of BCI systems. To evaluate the efficacy of channel selection for pediatric users, we tested four wrapper-based channel selection algorithms, sequential forward selection (SFS), sequential backward selection (SBS), sequential forward floating selection (SFFS), and sequential backward floating selection (SBFS) on offline motor imagery BCI data from three datasets involving typically developing children. The purpose was to assess the performance benefits and computational costs of each algorithm. All algorithms provided classification accuracy gains of 10–15 % with their optimal subsets. The time required to reach the optimal subsets varied between algorithms, but all took less than 80 s with mean completion times of 9.5 s and 35.8 s for the fastest (SFS) and slowest (SFFS), respectively. Adjusting the stopping criterion of the algorithm enables users to further reduce computation time with a disproportionately small effect on classification accuracy. All methods demonstrated an ability to prioritize expected physiological regions of interest and leave out channels detrimental to the classifier. Channel selection offers personalization of the BCI system for a specific user and a specific classifier. These findings emphasize the value of using personalized channel selection algorithms to improve motor imagery BCI systems for pediatric users.

Reinforcement Learning (RL) algorithms have recently been proposed to solve dynamic radio resource management (RRM) problems in beyond 5G networks. However, RL-based solutions are still not widely adopted in commercial cellular networks. One of the primary reasons for this is the slow convergence of RL agents when they are deployed in a live network and when the network’s context changes significantly. Concurrently, the open radio access network (O-RAN) paradigm promises to give mobile network operators (MNOs) more control over their networks, furthering the need for intelligent and RL-based network management. O-RAN’s standardized interfaces will allow MNOs to make real-time custom changes to intelligently control various RRM functionalities. We consider a RAN slicing scenario in which MNOs can modify the weights of the RL reward function. This enables MNOs to change the priorities of fulfilling the service level agreements of the slices. However, this results in a practical challenge since the RL agent needs to adapt promptly to the changes made by the MNO. This challenge is addressed in this paper, where we first present and discuss the results from an exhaustive experiment to examine the efficiency of using transfer learning (TL) to accelerate the convergence of RL-based RAN slicing in the considered scenario. We then propose a novelpredictiveapproach to enhance the TL-based acceleration by selecting the best-saved policy for reuse. By adopting the proposed policy transfer approach, RL agents are able to converge up to 14000 learning steps faster than their non-accelerated counterparts. The proposed machine learning (ML)-basedpredictiveapproach also shows up to a 96.5% accuracy in selecting the best expert policy to reuse for acceleration.

We consider online convex optimization (OCO) with multi-slot feedback delay. An agent selects a sequence of online decisions to minimize the accumulation of time-varying convex loss functions, subject to short-term and long-term constraints that may be time-varying. Both the convex loss function and the long-term constraint function may experience multiple time slots of feedback delay to be received by the agent. Existing works on OCO under this general setting has focused on the static regret, which measures the gap of losses between an online decision sequence and a time-invariant static offline benchmark. In this work, besides the static regret, we also consider a more practically meaningful metric, the dynamic regret, where the benchmark is a time-varying online optimal decision sequence. We propose an efficient algorithm, termed Delay-Tolerant Constrained-OCO (DTC-OCO), which uses a novel double regularization together with a new penalty mechanism on the long-term constraint violation, to tackle the asynchrony between information feedback and decision updates. We obtain upper bounds for its static regret, dynamic regret, and constraint violation, proving that they are sublinear under mild conditions. Furthermore, we consider a variation of DTC-OCO with multi-step gradient descent, and show it provides improved dynamic regret and constraint violation bounds for strongly convex loss functions. For numerical demonstration, we apply DTC-OCO to a general network resource allocation problem. Our simulation results suggest substantial performance gain by DTC-OCO over the current best alternative.

Cellular networks have evolved to deliver high-speed broadband services to support the requirements of IoT applications, which demand high speed, low latency, and massive capacity. A primary market goal is to provide support for ultra-reliable low latency communication (URLLC). URLLC requires sub-milliseconds-level latencies as defined by the third generation partnership project (3GPP). One of the promising technologies to achieve the aforementioned specifications is grant-free (GF) access for uplink resources. The GF scheme enables the user equipment (UE) to transmit data over pre-allocated resources which reduces communication latency. This paper proposes an intelligent Reinforcement Learning (RL) based allocator of grants trained via Deep Q-Learning. The experimental results show effect of the number of UEs in the network, and the percentage of unstable UEs on the speed of the RL agent's convergence.

Class-of-service (CoS) network traffic classification (NTC) classifies a group of similar traffic applications. The CoS classification is advantageous in resource scheduling for Internet service providers and avoids the necessity of remodelling. Our goal is to find a robust, lightweight, and fast-converging CoS classifier that uses fewer data in modelling and does not require specialized tools in feature extraction. The commonality of statistical features among the network flow segments motivates us to propose novel segmented learning that includes essential vector representation and a simple-segment method of classification. We represent the segmented traffic in the vector form using the essential vector representation (EVR). Then, the segmented traffic is modelled for classification using random forest based simple-segment method of classification (S2MC). Our solution's success relies on finding the optimal segment size and a minimum number of segments required in modelling. The solution is validated on multiple datasets for various CoS services, including virtual reality (VR). Significant findings of the research work are i) Synchronous services that require acknowledgment and request to continue communication are classified with 99 % accuracy, ii) Initial 1,000 packets in any session are good enough to model a CoS traffic for promising results, and we therefore can quickly deploy a CoS classifier, and iii) Test results remain consistent even when trained on one dataset and tested on a different dataset. In summary, our solution is the first to propose segmentation learning NTC that uses fewer features to classify most CoS traffic with an accuracy of 99 %. The implementation of our solution is available on GitHub.

Cooperative driving is a promising technology in the future Connected Autonomous Vehicles (CAV) because of its benefits to safety and fuel efficiency. However, since CAV will be relying heavily on wireless communication to cooperatively coordinate road maneuvering, latency and reliability of communication still pose a challenge. In this paper, we propose a novel scheme based on deep learning prediction to enhance the uplink resource allocation process in 5G C-V2X. The proposed scheme enables the base station to predict vehicle maneuvers, subsequently, assign it the required resource in advance without the need for scheduling request and granting process. This scheme improved the ability of 5G NR to support cooperative driving requirements. Moreover, we compare both traditional and proposed schemes discussing issues that arise from the introduction of prediction models and possible approaches for further enhancements in the future.

We consider federated learning in a wireless edge network, where multiple power-limited mobile devices collaboratively train a global model, using their local data with the assistance of an edge server. Exploiting over-the-air computation, the edge server updates the global model via analog aggregation of the local models over noisy wireless fading channels. Unlike existing works that separately optimize computation and communication at each step of the learning algorithm, in this work, we jointly optimize the training of the global model and the analog aggregation of local models over time. Our objective is to minimize the accumulated training loss at the edge server, subject to individual long-term transmit power constraints at the mobile devices. We propose an efficient algorithm, termed Online Model Updating with Analog Aggregation (OMUAA), to adaptively update the local and global models based on the time-varying communication environment. The trained model of OMUAA is channel- and power-aware, and it is in closed form with low computational complexity. We study the mutual impact between model training and analog aggregation over time, to derive performance bounds on the computation and communication performance metrics. Simulation results based on real-world image classification datasets and typical Long-Term Evolution network settings demonstrate substantial performance gain of OMUAA over the known best alternatives.

We consider cooperative multiple-input multiple-output (MIMO) precoding design with multiple access points (APs) assisted by a central controller (CC) in a fading environment. Even though each AP may have its own local channel state information (CSI), due to the communication delay in the backhaul, neither the APs nor the CC has timely global CSI. Under this semi-online setting, our goal is to minimize the accumulated precoding deviation between the actual local precoders executed by the APs and an ideal cooperative precoder based on the global CSI, subject to per-AP transmit power limits. We propose an efficient algorithm, termed Semi-Online Precoding with Information Parsing (SOPIP), which accounts for the network heterogeneity in information timeliness and computational capacity. SOPIP does not require the CC to send the full global CSI to each AP. Instead, it takes advantage of the precoder structure to substantially lower the communication overhead, while allowing each AP to effectively combine its own timely local CSI with the delayed global CSI to enable adaptive precoder updates. We analyze the performance of SOPIP in the presence of both multi-slot communication delay and gradient estimation error, showing that it has a bounded performance gap from an offline optimal solution. Simulation results under typical Long-Term Evolution network settings further demonstrate the substantial performance gain of SOPIP over other centralized and distributed schemes.

As one of the key technologies in 5G networks, Carrier Aggregation (CA) is studied in this paper. In CA, Component Carriers (CCs) can be activated and deactivated depending on multiple factors, e.g., energy consumption and Quality of Service (QoS) demand of users. We propose CC management strategies where each User Equipment (UE) minimizes its average delay and at the same time minimizes its power consumption while considering that CCs can be activated and deactivated only at certain times, as in real-world CA implementations. We first model the problem as a centralized multi-objective optimum CC management problem. Since centralized approaches would impose a large overhead on the system, we then develop a semi-distributed solution by modeling the problem as a stochastic game and propose a multi-agent Double Deep Q-Network (DDQN) based CC management algorithm to solve the stochastic game. We finally compare the proposed approaches with single CC activation and all-CC activation baseline schemes. Simulation results show that our proposed algorithms outperform the all-CC algorithm in terms of UE power consumption and have the capability of transmitting a number of bits with delay close to the all-CC scheme. Meanwhile, our DDQN-based algorithm decreases the UE power consumption by about 20% with respect to the all-CC scheme.

This paper presents a new virtualization method for the downlink of a multi-cell multiple-input multiple-output (MIMO) network, to achieve service isolation among multiple Service Providers (SPs) that share the base station resources of an Infrastructure Provider (InP). Each SP designs a virtual precoder for its users in each cell, as its service demand to the InP, without the need to be aware of the existence of the other SPs or to know the channel state information (CSI) outside the cell. The InP performs network virtualization to meet the SPs’ service demands while managing both the inter-SP and inter-cell interference. We consider coordinated multi-cell precoding at the InP and formulate an optimization problem to minimize a weighted sum of signal leakage and precoding deviation, with per-cell transmit power constraints. We propose a fully distributed semi-closed-form solution at each cell, without any CSI exchange across cells. We further propose a low-complexity scheme to allocate the virtual transmit power, for the InP to regulate between interference elimination and virtual demand maximization. Simulation results demonstrate that our precoding solution for network virtualization substantially outperforms the traditional spectrum isolation alternative. It can approach the performance of fully cooperative precoding when the number of antennas is large.

Connected Autonomous Vehicles (CAV) utilize vehicular communication to collect information about the surrounding environment to make informed decisions about speed and maneuvering. This enables safe driving and decreases the number of accidents and thereby the associated fatalities. However, vehicular communication may suffer from high latency and low reliability, especially in dense vehicle environments, which may negatively affect the safety of CAVs. Therefore, it is crucial to study the impact of these metrics on the safety application performance while taking into account realistic CAV kinematics and dynamics. In this paper, we address this problem by comparing the performance of the Short Range Communication (DSRC) to that of the Fifth-Generation New Radio (5G-NR) and their impacts on the safety applications in the CAV environment under different settings. We develop a full-fledged simulation framework that can realistically model both vehicular mobility and communication and can capture the impact of communication on safety applications. Within this framework, we implement an important CAV's safety application, namely, the forward collision avoidance system, in which following vehicles use vehicular communications to gather information from leading vehicles to compute the safe speed and avoid collisions. We then use this framework to study and compare the performance safety of the forward collision avoidance system using both DSRC and 5G-NR communications. The results show that the packet delays and drops in communication networks can adversely affect CAV safety. The results also demonstrate that 5G is more capable of supporting the safety requirements under higher packet traffic loads and vehicle densities.

Carrier aggregation (CA) is considered a key enabler technology for delivering higher rates to users of LTE and 5G networks. However, the increased transmission rate comes with the price of higher energy consumption which stems from users continuously monitoring the control channel of the active component carriers (CCs) whether data transmission is ongoing or not. In order to reduce energy consumption, we exploit the activation-deactivation procedure at the medium access control (MAC) layer of LTE/5G network. In this paper, we propose a reinforcement learning-based algorithm to improve energy-efficiency by dynamically activating-deactivating secondary component carriers (SCCs) with awareness of the user traffic profiles. The proposed algorithm aims to predict the arrival of data and identify SCCs to activate for each user. In addition, a traffic splitting approach and an intelligent exploration strategy are proposed to balance users' load among CCs and improve the convergence of the algorithm, respectively. Results of the proposed algorithm are compared with three baseline algorithms. The first baseline always activates all CCs for each user, the second baseline activates one carrier only (i.e., the primary carrier) and the third baseline algorithm relies on a reactive method, where the activation-deactivation decision is performed after observing the arrival of data. Results show that Q-learning outperforms the baseline algorithms by achieving the highest sum throughput (and lowest average delay) with the lowest number of activated SCCs, which is obtained by learning to dynamically activate SCCs according to the traffic pattern. Hence, Q-learning is considered the most energy-efficient compared to the baseline algorithms.

Cloud virtual reality (VR) gaming traffic characteristics such as frame size, inter-arrival time, and latency need to be carefully studied as a first step toward scalable VR cloud service provisioning. To this end, in this paper we analyze the behavior of VR gaming traffic and Quality of Service (QoS) when VR rendering is conducted remotely in the cloud. We first build a VR testbed utilizing a cloud server, a commercial VR headset, and an off-the-shelf WiFi router. Using this testbed, we collect and process cloud VR gaming traffic data from different games under a number of network conditions and fixed and adaptive video encoding schemes. To analyze the application-level characteristics such as video frame size, frame inter-arrival time, frame loss and frame latency, we develop an interval threshold based identification method for video frames. Based on the frame identification results, we present two statistical models that capture the behaviour of the VR gaming video traffic. The models can be used by researchers and practitioners to generate VR traffic models for simulations and experiments - and are paramount in designing advanced radio resource management (RRM) and network optimization for cloud VR gaming services. To the best of the authors' knowledge, this is the first measurement study and analysis conducted using a commercial cloud VR gaming platform, and under both fixed and adaptive bitrate streaming. We make our VR traffic datasets publicly available for further research by the community.

Unmanned Aerial Vehicles (UAV) are expected to play a crucial role in the future of 5G and beyond. However, designing efficient routing protocols for UAV is challenging due to the mobility and energy constraints. This problem becomes harder in UAV-assisted 5G networks because of its impact on the time slot assignment for the uplink and downlink in Time Division Duplex (TDD) frame structure. Thus, in this paper, we propose a new optimum routing technique for UAV-assisted TTD 5G networks, the Optimized Load-Balancing Routing (OLBR). The optimum routing problem is formulated in such a way that the decision variables are used to compute the time slot assignment in the 5G connection between UAV nodes. The objective of the optimization model is to minimize the network-wide delay. By distributing traffic across different alternative routes, OLBR minimizes network congestion, resulting in shorter queuing delays. Such a load-balancing also decreases the possibility of node failure due to energy depletion. The proposed OLBR is compared to the shortest path routing using Monte Carlo simulation on two different network topologies at different network traffic loads. The simulation results show that the OLBR produces significant savings in network-wide packet delay compared to the shortest path.

Mobile Edge Computing (MEC) refers to the concept of placing computational capability at the edge of the network to reduce the latency in handling the client requests. The performance of an edge server is adversely affected when it is overloaded, especially if it crashes due to overload and causes service failures. In this paper, a solution to prevent node from getting overloaded is analyzed by introducing an admission control policy. An adaptive admission control policy based low complexity RL (Reinforcement Learning) SALMUT (Structure-Aware Learning for Multiple Thresholds) is validated using several scenarios mimicking real world deployments. This approach performs as well as to the state-of-the-art deep RL algorithms such as PPO (Proximal Policy Optimization) and A2C (Advantage Actor Critic), but requires an order of magnitude less time to train, and outputs easily interpretable policy.

Carrier Aggregation (CA) has been a breakthrough in LTE that led to increased throughput for users, and is still one of the key technologies in 5G that helps to enhance spectrum utilization. In CA, Component Carriers (CCs) are dynamically activated and deactivated depending on several performance factors. Optimal selection of CCs has been studied in the literature. However, the latency associated with activation and deactivation of CCs, control channel overhead for switching CCs, as well as the energy consumed for monitoring the active CCs have not been a part of the optimal CC selection problem. Nevertheless, those become stringent design constraints in practice. In this paper, we address optimal CC selection and resource allocation in 5G networks, where the above constraints are considered and the 5G network supports several service types with different 5G QoS Identifiers (5QI). The proposed optimum joint CC selection and Radio Resource Block (RB) allocation schemes maximize average throughput of users and satisfy QoS of users in terms of delay. In addition, the proposed schemes take CC activation and deactivation burden into consideration and aim to minimize the number of activations and deactivations. The simulation results demonstrate that our proposed solution outperforms the state of the art solution while satisfying the QoS requirements and creating close to 95.5% reduction on the number of CCs activations and deactivations.

The ability to accurately forecast network resource utilization is vital in next-generation wireless networks. Based on the predicted load, telecom operators can proactively allocate network resources in an efficient way. In this paper, we perform a thorough analysis of a cellular network downlink load dataset collected at millisecond resolution. We first evaluate various statistical metrics of the physical resource block (PRB) utilization data to investigate its predictability. Then, we develop deep learning-based models to forecast PRB utilization in radio access networks (RANs). In particular, we propose univariate and multivariate long short-term memory (LSTM) network-based architectures for the forecasting task and investigate the impact of various prediction horizons and history lengths. When predicting PRB utilization, our approach showed up to 49% improvement in the Coefficient of Determination (r2score) and 19.5% decrease in the Root Mean Square Error (RMSE) compared with the baseline methods used.

We consider online convex optimization (OCO) with multi-slot feedback delay, where an agent makes a sequence of online decisions to minimize the accumulation of time-varying convex loss functions, subject to short-term and long-term constraints that are possibly time-varying. The current convex loss function and the long-term constraint function are revealed to the agent only after the decision is made, and they may be delayed for multiple time slots. Existing work on OCO under this general setting has focused on the static regret, which measures the gap of losses between the online decision sequence and an offline benchmark that is fixed over time. In this work, we consider both the static regret and the more practically meaningful dynamic regret, where the benchmark is a time-varying sequence of per-slot optimizers. We propose an efficient algorithm, termed Delay-Tolerant Constrained-OCO (DTC-OCO), which uses a novel constraint penalty with double regularization to tackle the asynchrony between information feedback and decision updates. We derive upper bounds on its dynamic regret, static regret, and constraint violation, proving them to be sublinear under mild conditions. We further apply DTC-OCO to a general network resource allocation problem, which arises in many systems such as data networks and cloud computing. Simulation results demonstrate substantial performance gain of DTC-OCO over the known best alternative.

Deep Reinforcement Learning (DRL) algorithms have been recently proposed to solve dynamic Radio Resource Management (RRM) problems in 5G networks. However, the slow convergence experienced by traditional DRL agents puts many doubts on their practical adoption in cellular networks. In this paper, we first discuss the need to have accelerated DRL algorithms. We then analyze the exploration behavior of various state-of-the-art DRL algorithms for slice resource allocation, and compare it with the traditional 5G Radio Access Network (RAN) slicing baselines. Finally, we propose a transfer learning-accelerated DRL-based solution for slice resource allocation. In particular, we tackle the challenge of slow convergence by transferring the policy learned by a DRL agent at an expert base station (BS) to newly deployed agents at target learner BSs. Our approach shows a remarkable reduction in convergence time and a significant performance improvement compared with its non-accelerated counterparts when tested against multiple traffic load variations.

To meet the ever-increasing communication services with diverse requirements, situation-aware intelligent utilization of multi-dimensional communication resources is becoming essential. In this paper, considering a time-division-duplex downlink cellular scenario, a deep learning-based framework for multi-dimensional intelligent multiple access (MD-IMA) scheme is developed for beyond 5G and 6G wireless networks to meet the real-time and diverse quality of service (QoS) requirements by fully utilizing the available radio resources in heterogeneous domains. To achieve intelligent operation of MD-IMA, the proposed deep learning scheme is achieved based on the convergence of long short term memory (LSTM) and deep reinforcement learning (DRL). Specifically, an LSTM neural network is used to predict the long-term network dynamics and inference changes in QoS requirements of the MD-IMA. Meanwhile, a deterministic policy gradient (DDPG) algorithm, a model-free DRL technique, is adopted to optimize the multi-dimensional radio resource allocation in real-time by dynamically following the fluctuations of the network situation. With the aid of the DDPG algorithm, radio resource management for MD-IMA can be achieved efficiently with reduced processing latency as compared to the conventional model-based approaches. Furthermore, the effectiveness of our proposed deep learning framework for MD-IMA is validated through real-world cellular traffic data-sets. The experimental results demonstrate that the proposed scheme can outperform state-of-the-art algorithms.

Network slicing is a key paradigm in 5G and is expected to be inherited in future 6G networks for the concurrent provisioning of diverse quality of service (QoS). Unfortunately, effective slicing of Radio Access Networks (RAN) is still challenging due to time-varying network situations. This paper proposes a new intelligent RAN slicing strategy with two-layered control granularity, which aims at maximizing both the long-term QoS of services and spectrum efficiency (SE) of slices. The proposed method consists of an upper-level controller to ensure the QoS performance, which enforces loose control by performing adaptive slice configuration according to the long-term dynamics of service traffic. The lower-level controller is to improve SE of slices, by tightly scheduling radio resources to users at the small time-scale. To realize the proposed RAN slicing strategy, we propose a model-free deep reinforcement learning (DRL) framework, which is a hierarchical structure that collaboratively integrating the modified deep deterministic policy gradient (DDPG) and double deep-Q-network algorithm. Specifically, the lower-level control problem is a mixed-integer stochastic optimization problem with multiple constraints. This kind of problem is hard to be directly solved by the exiting DRL algorithms, since it involves searching for the solution in a vast set of mixed-integer action space, which will induce unbearable computational complexity. Thus, we propose a novel action space reducing approach, embedding the convex optimization tools into the DDPG algorithm, to speed up the lower-level control. Furthermore, simulation results confirm the effectiveness of our proposed intelligent RAN slicing scheme.

The ever-growing wireless applications and their diverse Quality of Service (QoS) requirements bring the challenge of tailored QoS provisioning with limited radio resources in future cellular networks. While resource constraint is ubiquitous, different communication equipment in cellular networks could experience very different constraints in the multi-dimensional resource domains. To achieve stringent yet diverse QoS with limited resources, a novel multi-dimensional intelligent multiple access (MD-IMA) scheme is proposed in this paper to exploit disparate resource constraints among heterogeneous equipment for 5G beyond and 6G networks. With the assist of real-time data analysis, real-time QoS requirements, and resource availability of the related equipment are first determined in the proposed MD-IMA. Based on this, multiple access (MA) scheme is then intelligently adapted accordingly for each equipment in multi-dimensional resource domain to maximize the overall system requirement with operational constraints. The resource allocation in the MD-IMA system is further formulated as an optimization problem. To solve this non-convexity optimization of high computational complexity, the overall optimization is divided into several sub-problems and a joint optimization algorithm is adopted. Simulation results demonstrate the system energy efficiency performance gain of proposed MD-IMA over traditional MA is around 15% - 18%.

The past decade has witnessed a staggering evolution in cellular networks. Mobile wireless technologies have undergone four distinct generations; from uncomplicated voice calls in the first generation to high-speed, low latency and video streaming in the fourth generation. The numerous services brought to the users by 4G network have caused an increasing load demand. This increasing demand in network usage has proven the necessity of further service enhancements, such as predictive resource allocation techniques and handover analysis. For these techniques to be deployed, network quality and performance analysis must be performed on real-world network data. Since throughput is a major indicator of the network's performance, throughput modelling and prediction can be utilized for analyzing network quality. In this paper, two approaches for throughput analysis are examined: classical machine learning and time series forecasting. For the first approach, various machine learning models were deployed for throughput prediction and our analysis showed that the random forest model achieved the highest prediction performance. For time series forecasting, statistical methods as well as deep learning architectures were used. The evaluation shows that the machine learning models had a higher throughput prediction performance than the time series forecasting techniques.

With the advancements in wireless network technologies over the past few decades and the deployment of 4G LTE networks, the capabilities and services provided to end-users have become seemingly endless. Users of smartphones utilize high-speed network services while commuting on public transit and hope to have a consistent, high-quality connection for the duration of their trip. Due to the massive load demand on cellular networks and frequent changes in the underlying radio channel, users often experience sudden unexpected variations in the connection quality. To overcome such variations and maintain a consistent connection, these variations need to be predicted before they occur. This can be accomplished by the spatio-temporal analysis of the different network quality parameters and the investigation of the main factors that affect the network's performance and QoS. To this end, we conducted a network survey via Kingston Transit in Kingston, Ontario, Canada. We used the Android network monitoring application G-NetTrack Pro to build a dataset of various client-side wireless network quality parameters. The dataset consists of 30 repeated public transit bus trips at three different times of the day, each lasting around one hour. In this paper, we describe the data collection process, present an analysis of the collected data, and investigate the effects of time and location on the network's measured throughput and signal strength. We made the collected data, including more than 190 thousand unique records, publicly available to researchers in a domain where open data is rare.

Today deployments of massive Internet of Things (IoT) applications are expected from 5G networks. A primary challenge however is designing scalable wireless resource management schemes that can adapt to the varying temporal and spatial demand of IoT applications. As such, intelligence-based solutions that are agile to, and are able to exploit IoT traffic patterns are emerging as key enablers for 5G IoT applications. For example, Predictive Resource Allocation (PRA) has been proposed in wireless network literature as a mechanism to provide significant energy-savings and Quality of Experience (QoE) gains by leveraging predictions of the user location. While the results are very promising, further research is needed to 1) model and handle the inherent uncertainty in the predicted rates of PRA, and 2) develop low-complexity solutions for practical adoption. This is the topic of this paper, where we present a credibility-based chance-constrained fuzzy programming solution for PRA that enables the operator to control the energy efficiency-QoE tradeoff for different users and services. We demonstrate the use of a Kalman Filter (KF) to adaptively model rate prediction uncertainty by modifying the limits of the fuzzy membership functions in real-time. Our simulation results indicate that the proposed credibility-based framework provides a low-complexity solution for robust PRA.

With the deployment of large scale antenna array in millimeter wave (mmWave) band, the resolution of beamforming has been dramatically improved. To reduce the long beam-training process using reference signal (RS) in codebook-based high resolution beamforming, hierarchical codebook is often used to reduce the number of beam-training symbols. However, the large beam-training overhead is still the bottleneck for overall system performance improvement in term of the true achievable data rate. In this paper, with the angle reciprocity in frequency duplex division (FDD) system, a neural network based line of sight path angle of arrival (LAoA) estimation algorithm is proposed for beam selection, in order to achieve the non-RS-aided codebook-based beamforming. To further achieve high accuracy LAoA estimation, two-step neural network models are designed to capture the relationship between the receiving signal and the corresponding LAoA. The numerical results show that the proposed algorithm outperforms the benchmark algorithm in terms of sum weighted data rate (SWR) and sum data rate (SR). In the low signal to noise ratio (SNR) environments with a couple of uplink signal snapshots, our algorithm also performs better than MUSIC based beam selection algorithm.

Predictive resource allocations (PRAs) have recently gained attention in wireless network literature due to their significant energy-savings and quality of service (QoS) gains. This enhanced performance was primarily demonstrated while assuming the perfect prediction of both mobility traces and anticipated channel rates. While the results are very promising, several technical challenges need to be overcome before PRAs can be practically adopted. Techniques that model the prediction uncertainty and provide probabilistic quality of service (QoS) guarantees are among such challenges. This differs from the traditional robust optimization of wireless resources, as PRAs use a time horizon with predicted demands and anticipated data rates. In this paper, we tackle this problem and present an energy-efficient stochastic PRAs framework that is robust to prediction uncertainty under generic error probability density functions. The framework is applied for video delivery, where the desired video demands are modeled as probabilistic chance constraints over the prediction time horizon, and a deterministic closed form is then derived based on the Bernstein approximation (BA). In addition to handling prediction uncertainty, mechanisms that track the variance of the channel in real-time are practically needed. Towards this end, we demonstrate how a particle filter (PF) can be adopted to effectively achieve this functionality. A low complexity guided heuristic algorithm is also integrated with the BA-based allocations, and particle filter (PF), to provide a real-time solution. Extensive numerical simulations using a standard compliant long term evolution system are then presented to examine the developed solutions under various operating conditions. Results indicate the ability of our framework to significantly reduce base station energy consumption while satisfying users' QoS under practical prediction uncertainty.

Predictive resource allocation (PRA) techniques that exploit knowledge of the future signal strength along roads have recently been recognized as promising approaches to save base station (BS) energy and improve user quality of service (QoS). Recent studies on human mobility patterns and wireless signal strength measurements along buses and trains have indeed supported the practical potential of PRA. An unresolved challenge, however, is modeling the uncertainty in the predictions, and developing real-time robust solutions that incorporate probabilistic QoS guarantees. This is of paramount importance in PRA due to the prediction time horizon that adds considerable complexity and increases the rate uncertainty in the problem. With these developments in mind, this paper addresses energy-efficient PRA applied to stored video streaming using chance constrained programming. The proposed solution incorporates: 1) uncertainty in predicted user rates; 2) a joint level of probabilistic constraint satisfaction over a time horizon; and 3) both optimal gradient-based and real-time guided heuristic solutions. Our framework fundamentally differs from previous PRA work in the literature where nonstochastic approaches with assumptions of perfect prediction were primarily used to demonstrate the potential energy savings and QoS gains. Numerical simulations based on a standard compliant long term evolution (LTE) system are provided to examine and compare the developed solution. Unlike existing energy-efficient PRA, the proposed framework achieves the desired QoS level under imperfect channel predictions. This robustness is attained without compromising the energy-efficiency compared to opportunistic schedulers, and thus supports PRA implementation in practice.

The promising energy saving and QoS gains of Predictive Resource Allocation (PRA) techniques have recently been recognized in the wireless network research community. These gains were primarily introduced in light of perfect prediction of both mobility traces and anticipated channel rates. However, under real world considerations of prediction errors, the reported gains cannot be guaranteed and further investigation is needed. In this paper, we demonstrate the practical potential of PRA by developing a robust, probabilistic framework that guarantees QoS satisfaction for video streaming under imperfect predictions, without compromising the energy saving gains. The proposed PRA framework uses chance-constrained programming to model video streaming QoS for all users during the foreseen time horizon. Closed form solutions are developed using the Gaussian and Bernstein approximations based on the channel statistical measures. Extensive numerical simulations using a standard compliant Long Term Evolution (LTE) system are presented to examine the developed solutions, for different user mobility scenarios and target QoS levels. The results demonstrate the various design trade-offs involved toward the practical deployment of predictive video streaming in future generation networks.

Emerging mobility-aware content delivery approaches are being proposed to cope with the increasing usage of data from vehicular users. The main idea is to forecast the user locations and associated link capacity, and then proactively counter service fluctuations in advance. For instance, a user that is heading towards low coverage can be prioritized and have video content prebuffered. While the reported gains are encouraging, the results are primarily based on assumptions of perfect prediction. Investigating the predictability of mobility and future signal variations is therefore imperative to evaluate the practical viability of such predictive content delivery paradigms. To this end, this paper presents a large-scale measurement study of 33 repeated trips along a 23.4 km bus route covering urban and sub-urban areas in Kingston, Canada. We provide a thorough analysis of the collected traces to investigate the effects of geographical area, time, forecasting window, and contextual factors such as signal lights and bus stops. The collected dataset can also be used in several other ways to further investigate and drive research in predictive vehicular content delivery.

Abstract In current cellular networks, schedulers allocate wireless channel resources to users based on instantaneous channel gains and short‐term moving averages of user rates and queue lengths. By using only such short‐term information, schedulers ignore the users' service history in previous cells and, thus, cannot guarantee long‐term quality of service (QoS) when users traverse multiple cells with varying load and capacity. In this paper, we propose a new long‐term lookback scheduling (LLS) framework, which extends conventional short‐term scheduling with long‐term (QoS) information from previously traversed cells. We demonstrate the application of (LLS) for common channel aware, as well as channel and queue‐aware schedulers. The developed long‐term schedulers also provide a controllable trade‐off between emphasizing the immediate user (QoS) or the long‐term measures. Our simulation results show high gains in long‐term (QoS) without sacrificing short‐term user requirements. Therefore, the proposed scheduling approach improves subscriber satisfaction and increases operational efficiency. Copyright © 2014 John Wiley & Sons, Ltd.

Developing novel video delivery mechanisms have become imperative to cope with the unprecedented growth in mobile video traffic. In this paper, we present video transmission schemes that improve the streaming experience by looking ahead at the future rates users are expected to face. Such an approach is useful for the delivery of stored videos that can be strategically buffered in advance at the users' devices. For instance, if it is known a user is entering a low coverage area, content can be prebuffered to support smooth streaming. Therefore, the Base Stations (BSs) can now plan long-term multi-user rate allocations based not only on current channel states, but also on future conditions. To provide a performance benchmark we first develop a lookahead multi-objective Linear Program (LP) that offers a trade-off between minimizing overall network video degradation, and providing fairness in individual user degradation. Then, to efficiently solve the problem, we present a polynomial-time algorithm that closely follows the pareto-optimal trade-off of the multi-objective LP. We provide an extensive performance analysis of the proposed methods by simulations, and numerical results demonstrate that significant improvements in video streaming are achievable by the lookahead rate allocation strategies.

Novel mobility-aware resource allocation schemes have recently been introduced for efficient transmission of stored videos. The essence of such mechanisms is to lookahead at the future rates users will experience, and then strategically buffer content into user devices when they are at peak radio conditions. For example, a user approaching poor coverage will be preallocated additional video segments to ensure smooth streaming. Advances in mobility prediction and real-time radio environment map updates are driving forces for such Predictive Video Streaming (PVS) mechanisms. Although previous efforts have demonstrated the large potential gains of PVS, ideal channel predictions were assumed. This paper addresses the problem of channel uncertainty in PVS, and proposes a robust resource allocation framework that 1) models channel uncertainty, 2) solves the PVS problem with a tunable level of quality of service guarantees, and 3) learns the degree of uncertainty, and adapts the channel model accordingly. Numerical results demonstrate the effectiveness of the proposed approach for PVS under channel variability.

Small cell deployments have proven to be a cost-effective solution to meet the ever growing capacity and coverage requirements of mobile networks. While small cells are commonly deployed indoors, more recently outdoor roll-outs have garnered industry interest to complement existing macrocell infrastructure. However, the problem of where and when to deploy these small cells remains a challenge. In this paper, we investigate the small base station (SBS) placement problem in high demand outdoor environments. First, we propose a dynamic placement strategy (DPS) that optimizes SBS deployment for two different network objectives: minimizing data delivery cost, and minimizing macrocell utilization. We formulate each problem as a mixed integer linear program (MILP) that determines the optimal set of deployment locations among the candidate hot-spots to meet each network objective. Then we develop two greedy algorithms, one for each objective, that achieve close to optimal MILP performance. Our simulation results demonstrate that significant delivery cost and MBS utilization reductions are possible by incorporating the proposed deployment strategies.

Novel radio access techniques that leverage mobility predictions are receiving increasing interest in recent literature. The essence of these schemes is to lookahead at the future rates users will experience, and then devise long-term resource allocation strategies. For instance, a YouTube video user moving towards the cell edge can be prioritized to pre-buffer additional video content before poor coverage commences. While the potential of mobility-aware resource allocation has recently been demonstrated, several practical design aspects and evaluation approaches have not yet been addressed due to the complexity of the problem. Furthermore, since prior works have focused on specific applications there is also a strong need for a unified framework that can support different user and network requirements. For this purpose, we present a novel two-stage Predictive Radio Access Network (P-RAN) framework that can efficiently leverage both future data rate predictions in the order of tens of seconds, and instantaneous fast fading at the millisecond level. We also show how the framework can be implemented within the open source Network Simulator 3 (ns-3) LTE module, and apply it to optimize stored video delivery. A thorough set of performance tests are then conducted to assess the performance gains and investigate sensitivity to various prediction errors. Our results indicate that P-RANs can jointly improve both service quality and transmission efficiency. Additionally, we also observe that P-RAN performance can be further improved by modeling prediction uncertainty and developing robust allocation techniques.

Novel transmission mechanisms are imperatively needed to cope with the exponential growth of mobile traffic and its associated power consumption. To address such challenges, we present lookahead video delivery schemes that jointly improve the streaming experience and reduce BS power consumption. This is accomplished by exploiting knowledge of future wireless rates users are anticipated to face. Such an approach is useful for delivering stored videos that can be strategically buffered in advance at the users' devices. For instance, a user leaving the cell center may have content prebuffered efficiently before poor channel conditions prevail. This will save energy as transmission will not be needed during poor conditions. In this paper, we first formulate Lookahead Resource Allocation (LRA) as a multi-cell optimization problem that leverages predictions of user mobility rates. Then, we present centralized and distributed algorithms that closely follow the benchmark results of the optimal solution. Numerical results demonstrate that significant improvements in video streaming and BS power consumption are achievable by the LRA strategies.

Public sensing services utilizing the abundant on-vehicle resources are gaining high interest nowadays. One of the challenges facing such ubiquitous utilization is the recruitment and selection of the participating vehicles. In this paper, we present an optimal reputation-aware, trajectory-based framework that handles recruitment of vehicles for public sensing. The framework considers the spatiotemporal availability of participants along with their reputation to select vehicles that achieve desired coverage of an area of interest within a budget limit. In addition, we present a reputation assessment scheme and a pricing model for computing a reputation score and a recruitment cost for each candidate participant. The framework is formulated as an integer linear programming optimization problem and hence provides a benchmark and upper bound on achievable potential. We present analysis for two different practical recruitment objectives and show results under various scenarios.

In current cellular networks, schedulers allocate wireless channel resources to users based on short-term moving averages of the channel gain and of the queuing state. Using only such short-term information, schedulers ignore the user's service history in previous cells and, thus, cannot meet long-term Quality of Service (QoS) guarantees when users traverse cells with varying load and capacity. We propose a new scheduling framework, which extends conventional short-term scheduling with long-term QoS information from previously traversed cells. We demonstrate our scheme for relevant channel-aware as well as for channel and queue-aware schedulers. Our simulation results show high gains in long-term QoS while the average throughput of the network increases. Therefore, the proposed scheduling approach improves subscriber satisfaction while increasing operational efficiency.