A deep learning (DL)-based power control algorithm that solves the max-min user fairness problem in a cell-free massive multiple-input multiple-output (MIMO) system is proposed. Max-min rate optimization problem in a cell-free massive MIMO uplink setup is formulated, where user power allocations are optimized in order to maximize the minimum user rate. Instead of modeling the problem using mathematical optimization theory, and solving it with iterative algorithms, our proposed solution approach is using DL. Specifically, we model a deep neural network (DNN) and train it in an unsupervised manner to learn the optimum user power allocations which maximize the minimum user rate. This novel unsupervised learning-based approach does not require optimal power allocations to be known during model training as in previously used supervised learning techniques, hence it has a simpler and flexible model training stage. Numerical results show that the proposed DNN achieves a performance-complexity trade-off with around 400 times faster implementation and comparable performance to the optimization-based algorithm. An online learning stage is also introduced, which results in near-optimal performance with 4-6 times faster processing.

The problem of spectrum sharing between two operators in a dynamic network is considered. We allow both operators to share (a fraction of) their licensed spectrum band with each other by forming a common spectrum band. The objective is to maximize the gain in profits of both operators by sharing their licensed spectrum bands rather than using them exclusively, while considering the fairness among the operators. This is modeled as a two-person bargaining problem, and cast as a stochastic optimization. To solve this problem, we propose centralized and distributed dynamic control algorithms. At each time slot, the proposed algorithms perform the following tasks: 1) determine spectrum price for the operators; 2) make flow control decisions of users data; and 3) jointly allocate spectrum band to the operators and design transmit beamformers, which is known as resource allocation (RA). Since the RA problem is NP-hard, we have to rely on sequential convex programming to approximate its solution. To derive the distributed algorithm, we use alternating direction method of multipliers for solving the RA problem. Numerically, we show that the proposed distributed algorithm achieves almost the same performance as the centralized one. Furthermore, the results show that there is a trade-off between the achieved profits of the operators and the network congestion.Index Terms-Co-primary spectrum sharing, dynamic control, network utility maximization, stochastic optimization, Lyapunov drift, bargaining problem, fairness, sequential convex programming, alternating direction method of multipliers (ADMM), distributed algorithm.

Unmanned aerial vehicle (UAV) base stations (BSs) are reliable and efficient alternative to full fill the coverage and capacity requirements when the backbone network fails to provide such requirements due to disasters. In this paper, we consider optimal UAV-deployment problem in 3D space for a mmWave network. The objective is to deploy multiple aerial BSs simultaneously to completely serve the ground users. We develop a novel algorithm to find the feasible positions for a set of UAV-BSs from a predefined set of locations, subject to a signalto-interference-plus-noise ratio (SINR) constraint of every associated user, UAV-BS's limited hovering altitude constraint and restricted operating zone constraint. We cast this 3D positioning problem as an 0 minimization problem. This is a combinatorial, NP-hard problem. We approximate the 0 minimization problem as non-combinatorial 1-norm problem. Therefore, we provide a suboptimal algorithm to find a set of feasible locations for the UAV-BSs to operate. The analysis shows that the proposed algorithm achieves a set of the location to deploy multiple UVA-BSs simultaneously while satisfying the constraints.

Energy efficient beamforming and power control problem is considered for a MISO (multiple input single output) network. We consider an active femtocell within the coverage area of a macrocell. The femto base station is equipped with multi antennas and the users are considered to be single antenna users. A beamforming and power control problem is formulated in order to minimize the energy consumption per bit in the downlink transmission in the femtocell. The objective function is introduced as "sum power/sum rate" which has the unit J/bit to measure the energy efficiency of the network. The problem is nonconvex. We introduce a novel method to solve this problem with an approximation. We show that the problem can be solved with convex optimization techniques which has more practical interest even though the solution is suboptimal. In order to measure the energy efficiency we apply an existing power model which considers the total energy consumption of a base station. Thus we expect that our solution indicates realistic energy consumption measurements. Then we introduce a beamforming and power control algorithm which minimizes the energy consumption per bit transmission. Finally, the behavior of the objective function is observed in different antenna configurations by varying the user density in different channel environments.

Reconfigurable intelligent surface (RIS) is an emerging technology for improving performance in fifth-generation (5G) and beyond networks. Practically channel estimation of RISassisted systems is challenging due to the passive nature of the RIS. The purpose of this paper is to introduce a deep learningbased, low complexity channel estimator for the RIS-assisted multi-user single-input-multiple-output (SIMO) orthogonal frequency division multiplexing (OFDM) system with hardware impairments. We propose an untrained deep neural network (DNN) based on the deep image prior (DIP) network to denoise the effective channel of the system obtained from the conventional pilot-based least-square (LS) estimation and acquire a more accurate estimation. We have shown that our proposed method has high performance in terms of accuracy and low complexity compared to conventional methods. Further, we have shown that the proposed estimator is robust to interference caused by the hardware impairments at the transceiver and RIS.