In Internet of Things (IoT) networks, ensuring the timely delivery of information is significantly constrained by the limited energy resources of IoT devices and the signal attenuation experienced in wireless channels. In this paper, we investigate resource management for self-sustaining IoT networks with ambient radio frequency (RF) energy harvesting via a spatio-temporal approach. We consider a hard deadline for packet delivery, and we aim to jointly reduce the age of information (AoI) and the packet drop rate due to the hard deadline for packet delivery or buffer overflow. To achieve that, using tools from deep reinforcement learning (DRL) and stochastic geometry, we propose a joint rate adaptation and power control scheme that accounts for the spatial topology of the network and the temporal attributes at the device level. In particular, stochastic geometry is leveraged to characterize the energy harvesting process and the packet transmission success probability for a given transmit rate and power. Furthermore, the joint rate adaptation and power control policy at the device level is obtained using a deep R-network (DRN), which is a DRL algorithm that utilizes a deep neural network to approximate the R-function (the expected average reward). The performances of the last-come-first-served (LCFS) queuing discipline, the first-come-first-served (FCFS) queuing discipline, and a proposed hybrid queuing discipline are compared. For the proposed hybrid queuing discipline, DRL is used not only for rate adaptation and power control but also for specifying the transmission order of generated packets. The presented numerical results demonstrate that the LCFS queuing discipline improves AoI performance, while the FCFS queuing discipline improves packet drop rate. Also, the proposed hybrid queuing discipline strikes an intricate balance between AoI and packet drop rate, and achieves a good performance in both measures compared to the other queuing disciplines.

Multi-flow carrier aggregation (CA) is a key solution for improving data rates in heterogeneous networks (HetNets). Efficient data splitting is important to enhance the performance of multi-flow CA. In this paper, packet scheduling in multi-flow CA is investigated via a cross-layer design approach with comprehensive quality of service (QoS) provisioning. In particular, an approximation of the average queuing delay (AQD) in terms of the average buffers’ length and the average packet arrivals is presented. Then, using the presented AQD approximation, various packet scheduling schemes are developed in this paper. First, a low-complexity buffer and link aware heuristic packet scheduling scheme is proposed. Next, in order to reduce backhaul signalling overhead, a hybrid packet scheduling scheme that combines heuristic packet scheduling and random packet scheduling is considered. Then, an optimal packet scheduling scheme, albeit of high computational complexity, is proposed as a benchmark for other packet scheduling schemes. Also, the performances of the proposed packet scheduling schemes are analyzed and various data link layer performance parameters such as AQD, packet loss probability (PLP) and statistical queuing delay bound (SDB) are taken into consideration. Moreover, the complexity and the signalling overhead of the proposed packet scheduling schemes are investigated. Presented numerical results demonstrate that the heuristic packet scheduling scheme can achieve a performance that is similar to the performance of the optimal packet scheduling scheme while overcoming complexity and feasibility issues of the optimal packet scheduling scheme. Also, the hybrid packet scheduling scheme reduces signalling overhead while reaping some of the benefits of using buffer and link aware packet scheduling.

Multi-flow carrier aggregation (CA) is a key solution for the growing demand for high data rates in wireless networks. Efficient packet scheduling is important to improve the performance of multi-flow CA. In this paper, buffer-aware packet scheduling in multi-flow CA is investigated. First, an approximation of the average queuing delay in relation to the average buffers' length and the average packet arrivals is presented. Then, a buffer-aware packet scheduling scheme that minimizes the average queuing delay based on the presented approximation is proposed. Next, previously developed queuing analytical models are extended to evaluate the performance of the proposed buffer-aware packet scheduling scheme. Finally, the performance of the proposed buffer-aware packet scheduling scheme is compared with the state-of-the-art packet scheduling in multi-flow CA. It is shown that significant improvement is achieved by the proposed buffer-aware packet scheduling scheme not only in the queuing delay performance but also in the packet loss probability (PLP) performance. All analytical results are validated through computer simulations.

In this paper, we consider a coordinated multipoint dynamic cell selection (DCS) transmission scheme for serving sleeping cell user equipments (UEs). According to this DCS scheme, packets of UEs in a sleeping cell are randomly forwarded to the potential active base stations (BSs) by the packet serving gateway and UEs in the sleeping cell dynamically select their serving BS from these active BSs. We model the system as a fork/join (F/J) queuing system and develop a cross-layer analytical model that considers the time varying nature of the channels, channel scheduling mechanism, partial channel quality information feedback, cell selection mechanism, bursty packet arrivals, and packet scheduling mechanism. The developed analytical model can be used to measure various packet-level performance parameters, such as packet loss probability (PLP) and queuing delay while accounting for out-of-sequence packet delivery. We validate the accuracy of the developed analytical model via simulations, and we compare the performance of the DCS scheme under consideration with the conventional fixed cell selection scheme and with the state-of-the-art DCS scheme. Presented numerical results show that the DCS scheme under consideration significantly improves the PLP performance. Queuing delay performance, on the other hand, depends on the system and operating parameters.

In this paper, we consider a coordinated multipoint (CoMP) dynamic cell selection (DCS) scheme in order to improve performances of sleeping cell users. According to this DCS, packets for user equipments (UEs) in a sleeping cell are randomly forwarded to the potential active base stations (BSs) by the packet serving gateway (PSG) and UEs in the sleeping cell dynamically select these active BSs to be served in each time slot. We develop an analytical model using the fork/join (F/J) queuing model that considers the time varying nature of the channels, the channel scheduling mechanism, partial channel quality feedback, bursty packet arrivals and the packet scheduling mechanism. The developed analytical model can be used to measure various packet level performance parameters such as packet loss probability (PLP) and queuing delay while accounting for out-of-order packet delivery. Presented numerical results show that the considered CoMP DCS scheme provides significant performance improvement compared to the traditional fixed cell selection.

In this paper, we adopt an energy-efficiency (EE) metric, namedworst-EE, that is suitable for EE fairness optimization in the uplink transmission of amplify-and-forward (AF) cooperative orthogonal frequency division multiple access (OFDMA) networks. More specifically, we assign subcarriers and allocate powers for mobile and relay stations in order to maximize the worst-EE, i.e., to maximize the EE of the mobile station (MS) with the lowest EE value, subject to MSs transmit power, relay station (RS) transmit power, and MSs quality-of-service (QoS) constraints. The formulated primal max–min optimization problem is nonconvex fractional mixed integer nonlinear program, i.e., NP-hard to solve. We provide a novel optimization framework that studies the structure of the primal problem and prove that the dual min–max optimization problem attains the same optimal solution of the primal problem. Additionally, we propose a modified Dinkelbach algorithm, named dual Dinkelbach, to achieve the optimal solution of the dual problem in a polynomial time complexity. We further exploit the structure of the obtained optimal solution and develop a low complexity suboptimal heuristic. Numerical results show the effectiveness of the proposed algorithm to improve the network performance in terms of fairness between MSs, worst-EE, and average network transmission rate when compared to traditional schemes that maximize the EE of the whole network. Presented results also show that the suboptimal heuristic balances the achieved performance and the computational complexity.

Small-cell networks (SCNs) have emerged as a potential solution to the rapidly increasing demand for high-data-rate services over wireless networks. The small-cell access nodes (SANs) provide service to users through the access link and can be connected to the core/global network, preferably, via a wireless backhaul link. In this paper, we develop a queuing analytical model that considers the channel scheduling mechanisms in both links of SCNs, the time-varying nature of the channels, bursty packet arrivals, and the network topology e.g., the number of SANs and the coverage of the small cells. For the access link, we consider the so-called max rate/opportunistic channel scheduling mechanism, while for the backhaul link, we consider three different channel scheduling mechanisms, namely, fixed channel scheduling, round-robin channel scheduling, and access-link-dependent channel scheduling. Our developed analytical model is useful for gauging various data-link-layer performance measures, such as packet loss probability and average queuing delay of the packets, for the channel scheduling mechanisms under consideration. Presented numerical examples show that the choice of channel scheduling mechanism in the backhaul link is not unique and it depends on the operating scenario and required quality-of-service (QoS) parameters. The developed queuing model can also facilitate cross-layer design to meet the required QoS parameters. Presented simulation results validate the accuracy of the developed model.

Small-cell networks (SCNs) have emerged as a potential solution to the rapidly increasing demand for high data rates. The base stations (also referred to as access nodes) of the small cells provide service to a group of users through the access link and they are connected to the core/global network via a backhaul link. In this paper, we develop a queuing analytical model that considers the channel scheduling mechanisms in both links, the time varying nature of the channels as well as bursty packet arrivals. For the access link we consider the so-called max rate/opportunistic channel scheduling mechanism, while for the backhaul link we consider three different channel scheduling mechanisms, namely, fixed channel scheduling, round-robin channel scheduling and an access-link dependent channel scheduling mechanisms. Our developed analytical model is useful to measure various data link layer performances such as packet loss probability and average queuing delay. These performance measures can be used to determine which of the scheduling mechanisms considered for the backhaul link provides the best performance.