Publications

In this paper, the secrecy outage performance of a multiple-input single-output (MISO) wiretap channel is investigated. Different from previous works, we assume that the legitimate transmitter employs maximal-ratio transmission (MRT) beamforming scheme with outdated channel state information (CSI) of the main channel. A closed-form expression for the secrecy outage probability is derived, from which an asymptotic secrecy analysis is carried out. Furthermore, a closed-form expression for the probability of non-zero secrecy rate is derived. Our results are corroborated by Monte Carlo simulations and show that full diversity gain cannot be realized in the presence of outdated CSI. In addition, the MRT scheme is analytically compared with a transmit antenna selection scheme and the secrecy outage improvement is quantified.

This paper delves into the unexplored frequency-dependent characteristics of beyond diagonal reconfigurable intelligent surfaces (BD-RISs). A generalized practical frequency-dependent reflection model is proposed as a fundamental framework for configuring fully-connected and group-connected RISs in a multi-band multi-base station (BS) multiple-input multiple-output (MIMO) network. Leveraging this practical model, multi-objective optimization strategies are formulated to maximize the received power at multiple users connected to different BSs, each operating under a distinct carrier frequency. By relying on matrix theory and exploiting the symmetric structure of the reflection matrices inherent to BD-RISs, closed-form relaxed solutions for the challenging optimization problems are derived. The ideal solutions are then combined with codebook-based approaches to configure the practical capacitance values for the BD-RISs. Simulation results reveal the frequency-dependent behaviors of different RIS architectures and demonstrate the effectiveness of the proposed schemes. Notably, BD-RISs exhibit superior resilience to frequency deviations compared to conventional single-connected RISs. Moreover, the proposed optimization approaches prove effective in enabling the targeted operation of BD-RISs across one or more carrier frequencies. The results also shed light on the potential for harmful interference in the absence of proper synchronization between RISs and adjacent BSs.

An reconfigurable intelligent surface (RIS) can be used to establish line-of-sight (LoS) communication when the direct path is compromised, which is a common occurrence in a millimeter wave (mmWave) network. In this paper, we focus on the uplink channel estimation of a such network. We formu-late this as a sparse signal recovery problem, by discretizing the angle of arrivals (AoAs) at the base station (BS). On-grid and off-grid AoAs are considered separately. In the on-grid case, we propose an algorithm to estimate the direct and RIS channels. Neural networks trained based on supervised learning is used to estimate the residual angles in the off-grid case, and the AoAs in both cases. Numerical results show the performance gains of the proposed algorithms in both cases.

This paper presents a framework to evaluate the probability that a decision error event occurs in wireless sensor networks, including sensing and communication errors. We consider a scenario where sensors need to identify whether a given event has occurred based on its periodic, noisy, observations of a given signal. Such information about the signal needs to be sent to a fusion center that decides about the actual state at that specific observation time. The communication links -- single- or multi-hop -- are modeled as binary symmetric channels, which may have different error probabilities. The decision at the fusion center is based on OR, AND, K-OUT-OF-N and MAJORITY Boolean operations on the received signals associated to individual sensor observations. We derive closed-form equations for the average decision error probability as a function of the system parameters (e.g. number of sensors and hops) and the input signal characterization. Our analyses show the best decision rule is closely related to the frequency that the observed events occur and the number of sensors. In our numerical example, we show that the AND rule outperforms MAJORITY if such an event is rare and there is only a handful number of sensors. Conversely, if there is a large number of sensors or more evenly distributed event occurrences, the MAJORITY is the best choice. We further show that, while the error probability using the MAJORITY rule asymptotically goes to 0 with increasing number of sensors, it is also more susceptible to higher channel error probabilities.

We consider a single-macrocell heterogeneous multiple-input multiple-output network, where the macrocell shares the same frequency band with the femto network. The interference power to the macro users from the femto base stations is kept below a threshold to guarantee that the performance of the macro users does not degrade due to the femto network. We consider the problem of finding the set of all achievable power-rate tuples for this setting. We first formulate a two-dimensional vector optimization problem in which we consider maximizing the sum-rate and minimizing the sum-power, subject to maximum power and interference threshold constraints. The considered problem is NP-hard. We provide a method to solve the problem by using the relationship between the weighted sum-rate maximization and weighted-sum-mean-squared-error minimization problems. Furthermore, using the proposed algorithm, we evaluate the impact of imposing interference threshold constraints and the impact of co-channel deployment in heterogeneous networks. The proposed algorithm can be used to evaluate the performance of real heterogeneous networks via off-line numerical simulations.

We consider a K link multiple-input multiple-output (MIMO) interference channel, where each link consists of two full-duplex (FD) nodes exchanging information simultaneously in a bi-directional communication fashion. The nodes in each pair suffer from self-interference due to operating in FD mode, and inter-user interference from other links due to simultaneous transmission at each link. We consider the transmit and receive filter design for weighted sum-rate (WSR) maximization problem subject to sum-power constraint of the system or individual power constraints at each node of the system. Based on the relationship between WSR and weighted minimum-mean-squared-error (WMMSE) problems for FD MIMO interference channels, we propose a low complexity alternating algorithm which converges to a local WSR optimum point. Moreover, we show that the proposed algorithm is not only applicable to FD MIMO interference channels, but also applicable to FD cellular systems in which a base station (BS) operating in FD mode serves multiple uplink (UL) and downlink (DL) users operating in half-duplex (HD) mode, simultaneously. It is shown in simulations that the sum-rate achieved by FD mode is higher than the sum-rate achieved by baseline HD schemes.

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.

Data aggregation is a promising approach to enable massive machine-type communication (mMTC). This paper focuses on the aggregation phase where a massive number of machine-type devices (MTDs) transmit to aggregators. By using non-orthogonal multiple access (NOMA) principles, we allow several MTDs to share the same orthogonal channel in our proposed hybrid access scheme. We develop an analytical framework based on stochastic geometry to investigate the system performance in terms of average success probability and average number of simultaneously served MTDs, under imperfect successive interference cancellation (SIC) at the aggregators, for two scheduling schemes: random resource scheduling (RRS) and channel-aware resource scheduling (CRS). We identify the power constraints on the MTDs sharing the same channel to attain a fair coexistence with purely orthogonal multiple access (OMA) setups, then power control coefficients are found so that these MTDs perform with similar reliability. We show that under high access demand, the hybrid scheme with CRS outperforms the OMA setup by simultaneously serving more MTDs with reduced power consumption.

Recent trends show that mobile users accessing the wireless communication network from public transportation vehicles such as the high speed train are on the increase. To serve on-board users effectively, the use of a two-hop communication link is a promising solution. However existing two-hop schemes are expensive technologies and not integral part of the cellular network. We focus on a two-hop integrated cellular link which consists of the base station-to-train link and the train-to-on board users link which is created with the use of moving relay nodes (MRNs). The base station-to-train link is considered in a high speed scenario. We propose an iterative algorithm that can be used to efficiently select the appropriate number of MRNs and receive antennas in relation to the throughput performance and capital expenses.

Channel state information (CSI) has been a key component in traditional wireless communication systems. This might no longer hold in future networks supporting services with stringent quality of service constraints such as extremely low-latency, low-energy and/or large number of simultaneously connected devices, where acquiring CSI would become extremely costly or even impractical. We overview the main limitations of CSI at transmitter side (CSIT)-based designs toward the 6G era, assess how to design and use efficient CSIT-limited schemes that allow meeting the new and stringent requirements, and highlight some key research directions. We delve into the ideas of efficiently allocating pilot sequences, relying on statistical CSIT and/or using location-based strategies, and demonstrate viability via a selected use case.

The architectures of mobile networks have seen an unprecedented techno-economic transformation, fusing the telcommunications world within the cloud world, adding the spices of Software Engineering to the overall system design, and ultimately yielding the concept of Telco Cloud. This has brought significant benefits in terms of reducing expenditure and operational costs, flexibility in deployment, and a faster time to market. The key enablers are network function virtualization, software-defined networking, and edge/cloud computing. Artificial intelligence is also kicking in this arena. When all these technologies are well integrated, the creation and life-cycle management of fully programmable, flexible, service-tailored, and automated end-to-end network slices/services become possible. This will support diverse 5G and beyond 5G services, spanning from tactile IoT to pervasive robotics and immersive services. This paper introduces an unprecedented and disruptive vision for 6G that shifts the perception of future mobile networks from the old-fashioned concept of "network of networks" towards a new vision of "service of services." The paper then introduces the functional model of the envisioned system architecture, along with its components. It then provides a high-level description of the logical architecture.

We study the joint active/passive beamforming and channel blocklength (CBL) allocation in a non-ideal reconfigurable intelligent surface (RIS)-aided ultra-reliable and low-latency communication (URLLC) system. The considered scenario is a finite blocklength (FBL) regime and the problem is solved by leveraging a novel deep reinforcement learning (DRL) algorithm named twin-delayed deep deterministic policy gradient (TD3). First, assuming an industrial automation system with multiple actuators, the signal-to-interference-plus-noise ratio and achievable rate in the FBL regime are identified for each actuator in terms of the phase shift configuration matrix at the RIS. Next, the joint active/passive beamforming and CBL optimization problem is formulated where the objective is to maximize the total achievable FBL rate in all actuators, subject to non-linear amplitude response at the RIS elements, BS transmit power budget, and total available CBL. Since the amplitude response equality constraint is highly non-convex and non-linear, we resort to employing an actor-critic policy gradient DRL algorithm based on TD3. The considered method relies on interacting RIS with the industrial automation environment by taking actions which are the phase shifts at the RIS elements, CBL variables, and BS beamforming to maximize the expected observed reward, i.e., the total FBL rate. We assess the performance loss of the system when the RIS is non-ideal, i.e., with non-linear amplitude response, and compare it with ideal RIS without impairments. The numerical results show that optimizing the RIS phase shifts, BS beamforming, and CBL variables via the proposed TD3 method is highly beneficial to improving the network total FBL rate as the proposed method with deterministic policy outperforms conventional methods.

In this paper, the average achievable rate and error probability of a reconfigurable intelligent surface (RIS) aided systems is investigated for the finite blocklength (FBL) regime. The performance loss due to the presence of phase errors arising from limited quantization levels as well as hardware impairments at the RIS elements is also discussed. First, the composite channel containing the direct path plus the product of reflected channels through the RIS is characterized. Then, the distribution of the received signal-to-noise ratio (SNR) is matched to a Gamma random variable whose parameters depend on the total number of RIS elements, phase errors and the channels' path loss. Next, by considering the FBL regime, the achievable rate expression and error probability are identified and the corresponding average rate and average error probability are elaborated based on the proposed SNR distribution. Furthermore, the impact of the presence of phase error due to either limited quantization levels or hardware impairments on the average rate and error probability is discussed. The numerical results show that Monte Carlo simulations conform to matched Gamma distribution to received SNR for sufficiently large number of RIS elements. In addition, the system reliability indicated by the tightness of the SNR distribution increases when RIS is leveraged particularly when only the reflected channel exists. This highlights the advantages of RIS-aided communications for ultra-reliable and low-latency systems. The difference between Shannon capacity and achievable rate in FBL regime is also discussed. Additionally, the required number of RIS elements to achieve a desired error probability in the FBL regime will be significantly reduced when the phase shifts are performed without error.

In this paper, an adaptive scheme for amplify-and-forward relaying networks is proposed, which selects a certain transmission mode for each communication process. Depending on the instantaneous channel conditions, one of the following modes is selected: direct transmission with no cooperation, cooperative transmission with half-duplex relaying and maximal-ratio combining at the destination, or cooperative transmission with full-duplex relaying and maximal-ratio combining at the destination. A three-node network is considered, containing a single-antenna source, a two-antenna relay that is able to implement full-duplex communication, and a single-antenna destination. Energy normalization per block is assumed, so that in those modes using cooperation, the system's transmission power is shared between source and relay. The performance analysis is provided in terms of outage probability and energy efficiency. We derive a tight approximate expression in closed form for the outage probability and an approximate expression in integral form for the mean energy consumption. The results show that our scheme outperforms all of transmission modes separately in terms of outage probability, while being more energy efficient than the cooperative transmission modes. In addition, the asymptotic analysis proves that the proposed scheme achieves full diversity order equal to 2, thus outperforming those schemes with direct transmission or full-duplex cooperation only.

Virtualized Radio Access Networks (vRANs) are fully configurable and can be implemented at a low cost over commodity platforms to enable network management flexibility. In this paper, a novel vRAN reconfiguration problem is formulated to jointly reconfigure the functional splits of the base stations (BSs), locations of the virtualized central units (vCUs) and distributed units (vDUs), their resources, and the routing for each BS data flow. The objective is to minimize the long-term total network operation cost while adapting to the varying traffic demands and resource availability. Testbed measurements are performed to study the relationship between the traffic demands and computing resources, which reveals high variance and depends on the platform and its load. Consequently, finding the perfect model of the underlying system is non-trivial. Therefore, to solve the proposed problem, a deep reinforcement learning (RL)-based framework is proposed and developed using model-free RL approaches. Moreover, the problem consists of multiple BSs sharing the same resources, which results in a multi-dimensional discrete action space and leads to a combinatorial number of possible actions. To overcome this curse of dimensionality, action branching architecture, which is an action decomposition method with a shared decision module followed by neural network is combined with Dueling Double Deep Q-network (D3QN) algorithm. Simulations are carried out using an O-RAN compliant model and real traces of the testbed. Our numerical results show that the proposed framework successfully learns the optimal policy that adaptively selects the vRAN configurations, where its learning convergence can be further expedited through transfer learning even in different vRAN systems. It offers significant cost savings by up to 59\% of a static benchmark, 35\% of DDPG with discretization, and 76\% of non-branching D3QN.