While machine-type communication (MTC) devices generate considerable amounts of data, they often cannot process the data due to limited energy and computational power. To empower MTC with intelligence, edge machine learning has been proposed. However, power allocation in this paradigm requires maximizing the learning performance instead of the communication throughput, for which the celebrated water-filling and max-min fairness algorithms become inefficient. To this end, this paper proposes learning centric power allocation (LCPA), which provides a new perspective on radio resource allocation in learning driven scenarios. By employing 1) an empirical classification error model that is supported by learning theory and 2) an uncertainty sampling method that accounts for different distributions at users, LCPA is formulated as a nonconvex nonsmooth optimization problem, and is solved using a majorization minimization (MM) framework. To get deeper insights into LCPA, asymptotic analysis shows that the transmit powers are inversely proportional to the channel gains, and scale exponentially with the learning parameters. This is in contrast to traditional power allocations where quality of wireless channels is the only consideration. Last but not least, a large-scale optimization algorithm termed mirror-prox LCPA is further proposed to enable LCPA in large-scale settings. Extensive numerical results demonstrate that the proposed LCPA algorithms outperform traditional power allocation algorithms, and the large-scale optimization algorithm reduces the computation time by orders of magnitude compared with MM-based LCPA but still achieves competing learning performance.

With the development of the maritime networks, the data of vessel users is growing exponentially, more and more resource intensive tasks, such as multimedia applications, high-definition video playback and games, appear in the daily demands. These changes have greatly increased the energy consumption and bandwidth requirements of vessel terminals and networks. In order to meet the requirements of high bandwidth and low delay for the high-speed development of mobile network, and reduce the network load, the concept of mobile edge computing (MEC) is proposed and has been widely supported by the academia and industry. It is considered to be one of the key technologies of the next generation networks. Inspired by this idea, this paper introduces computing offloading technology to maritime mobile cloud networks. Maritime mobile cloud network is the product of the continuous development of cloud computing technology and mobile Internet technology. In this article, we studied the issue of computation task offloading for vessel terminals, focusing on minimizing the energy consumption of vessel terminals and the execution delay of computation task. Firstly, it determines that whether if it should be offloaded to the cloud server. Secondly, the server should be selected to run the computation task. The goal of the optimization is to minimize the energy consumption of vessel terminals and the execution delay of computation task taking into account of different weights. To reduce the execution latency and device energy consumption, we proposed a multi-vessel computation offloading algorithm based on improved Hungarian algorithm in maritime mobile edge computing network. Finally, simulation results demonstrate the effectiveness of the proposed scheme.

Mobile edge computing (MEC) has emerged in recent years as an effective solution to the challenge of limited vehicle resources in the Internet of Vehicles (IoV), especially for computation-intensive vehicle tasks. This paper investigates a multi-user MEC system with an active task model in high-dynamic IoV scenarios. To improve the MEC performance regarding system capacity, task service delay, and energy consumption, we design an adaptive joint resource allocation scheme based on deep reinforcement learning (DRL), which includes uplink, computing, and downlink resource allocation. Further, a multi-actor parallel twin delayed deep deterministic policy gradient (MAPTD3) algorithm is devised to jointly and adaptively optimize these strategies during each time slot. Finally, numerical results demonstrate that the proposed adaptive joint resource allocation scheme improves system performance significantly while satisfying task delay and system resource constraints. In addition, the space complexity of the designed optimization algorithm is lower than that of conventional DRL algorithms.

This work investigates the energy-efficient resource allocation for layered-division multiplexing (LDM) based non-orthogonal multicast and unicast transmission in cell-free massive multiple-input multiple-output (MIMO) systems, where each user equipment (UE) performs wireless information and power transfer simultaneously. To begin with, the achievable data rates for multicast and unicast services are derived in closed form, as well as the received radio frequency (RF) power at each UE. Based on the analytical results, a nonsmooth and nonconvex optimization problem for energy efficiency (EE) maximization is formulated, which is however a challenging fractional programming problem with complex constraints. To suit the massive access setting, a first-order algorithm is developed to find both initial feasible point and the nearly optimal solution. Moreover, an accelerated algorithm is designed to improve the convergence speed. Numerical results demonstrate that the proposed first-order algorithms can achieve almost the same EE as that of second-order approaches yet with much lower computational complexity, which provides insight into the superiority of the proposed algorithms for massive access in cell-free massive MIMO systems.

This article performs a unified analysis of three coordinated multipoint (CoMP) transmission strategies in the downlink of mmWave cellular networks, including the fixed-number base station (BS) cooperation (FNC), the fixed-region BS cooperation (FRC), and the interference-aware BS cooperation (IAC). We first develop a comprehensive framework for CoMP operation in cellular networks, and investigate the network performance under a Poisson point process (PPP) model together with mmWave spectrum. To show what fraction of users in the network achieve target reliability for a given signal to interference-plus-noise ratio (SINR)/signal-to-interference ratio (SIR), we derive the SINR/SIR meta distributions, and further obtain the coverage probability as well as mean local delay for the three cooperation strategies. A pivotal intermediate step to compute the performance metrics is the derivation of joint distributions of distances between a typical user and cooperative BSs. Our analysis demonstrates that parameters of blockage have a significant influence on the network performance for the three CoMP schemes. We find that the FRC scheme makes more users achieve the given link reliability for the scenario with a low density of BSs, while the IAC scheme provides better performance for the network with a high density of BSs. Moreover, the optimal CoMP scheme can be approximately selected by considering the nearest distance from the serving BS to user and the radius of the approximate line-of-sight (LoS) region in the cellular networks.

In this paper, a novel 3D cellular model consisting of aerial base stations (aBSs) and aerial user equipments (aUEs) is proposed, by integrating the coordinated multi-point (CoMP) transmission technique with the theory of stochastic geometry. For this new 3D architecture, a tractable model for aBSs’ deployment based on the binomial-Delaunay tetrahedralization is developed, which ensures seamless coverage for a given space. In addition, a versatile and practical frequency allocation scheme is designed to eliminate the inter-cell interference effectively. Based on this model, performance metrics including the achievable data rate and coverage probability are derived for two types of aUEs: i) the general aUE (i.e., an aUE having distinct distances from its serving aBSs) and ii) the worst-case aUE (i.e., an aUE having equal distances from its serving aBSs). Simulation and numerical results demonstrate that the proposed approach emphatically outperforms the conventional binomial-Voronoi tessellation without CoMP. Insightfully, it provides a similar performance to the binomial-Voronoi tessellation which utilizes the conventional CoMP scheme; yet, introducing a considerably reduced computational complexity and backhaul/signaling overhead.