DECT-2020 NR is a radio access technology designed for 5G Internet of Things (IoT) applications. By leveraging multi-hop communications and listen-before-talk (LBT) mechanisms, DECT-2020 NR offers a flexible and cost-effective deployment solution for massive machine-type communications (mMTC). The ETSI standardization committee currently considers extending use-case options to ultra-reliable low-latency communications by enabling scheduled access along with LBT. However, this approach can increase the cost of the end systems. The aim of this study is to investigate whether standardized physical and medium access control (MAC) mechanisms, including power control, LBT, and priority queuing, can provide near-URLLC operation over multihop topologies. Our results show that to improve loss performance one needs to disable power control, use a shorter back-off window, and enable priority data transmission. For improving latency performance it is critical to use shorter acknowledgement (ACK) timeout and send ACKs immediately after data packet reception. Applying these techniques allows us to improve multi-hop URLLC latency performance by up to 30-80% and loss performance by up to 10-20% depending on URLLC and mMTC traffic conditions with only negligible impact on mMTC traffic performance.

The Age of Information (AoI) is a metric for state update applications, which plays a crucial role in the Internet of Things (IoT). In many environments, such as factories and energy grid deployment, end sensing nodes tend to be naturally clustered, resulting in high variability and autocorrelation in the packet arrival process. In this paper, we assess the impact of variability and autocorrelation of packet arrivals in one cluster on the performance of other clusters in 5G cellular systems by explicitly accounting for the batch nature of the packet arrival and service processes arising as a a result of discrete nature of the scheduling process and the autocorrelational properties of the arrival processes. We show that not only the mean arrival rates but coefficient of variation (CoV) and autocorrelation in the arrival processes from tagged and competing clusters affect both the mean and distribution of the peak AoI (PAoI). Although the impact of these second-order properties depends heavily on their relative loads, it is tagged cluster properties that produces the most impact on its own performance metrics. Notably, the worst performance is observed when the competing traffic takes on the properties of the Poisson process – unit variability and no memory. The main practical takeaway is that increase in either tagged or background sources variability can raise the PAoI by 40-60%, but it also makes it more stable around the mean value reducing the jitter at the destination.

The swarms of unmanned aerial vehicles (UAV) are nowadays finding numerous applications in different fields. While performing their missions, UAVs have to rely on external positioning information to maintain connectivity and communications between units in a swarm. However, some of the critical applications such as rescue missions are performed in locations, where this information is partially or fully not available, e.g., deep woods, mountains, indoors. In this paper, we propose a method for dynamic topology organization and maintenance in UAV swarms. In addition to the baseline functionality, we also design advanced features required for dynamic swarms merging and disjoining, making it suitable for practical applications. Specifically, the proposal is based on the virtual coordinates system allowing for the utilization of conventional geographical routing algorithms. We test the proposed algorithm in different swarm conditions to illustrate that: (i) it is insensitive to distance estimates up to at least 30% allowing for simple estimation techniques, (ii) the accuracy of the topology inference is at least 90% even under impairments caused by mobility and temporal loss of connectivity, and (iii) the impact of the developed merging algorithm for swarms lasts for multiple tens of time steps that correspond to just few seconds in practice. The set of developed algorithms can be utilized to ensure always connected topology in conditions where positioning information is partially or fully unavailable.

Network slicing is a technique introduced by 3GPP to enable multi-tenant operation in 5 G systems. However, the support of slicing at the air interface requires not only efficient optimization algorithms operating in real time but also its tight integration into the 5 G control plane. In this paper, we first present a priority-based mechanism enabling defined performance isolation among slices competing for resources. Then, to speed up the resource arbitration process, we propose and compare several supervised machine learning (ML) techniques. We show how to embed the proposed approach into the ITU-T standardized ML architecture. The proposed ML enhancement is evaluated under realistic traffic conditions with respect to the performance criteria defined by GSMA while explicitly accounting for 5 G millimeter wave channel conditions. Our results show that ML techniques are able to provide suitable approximations for the resource allocation process ensuring slice performance isolation, efficient resource use, and fairness. Among the considered algorithms, polynomial regressions show the best results outperforming the exact solution algorithm by 5–6 orders of magnitude in terms of execution time and both neural network and random forest algorithms in terms of accuracy (by 20–40 %), sensitiveness to workload variations and training sample size. Finally, ML algorithms are generally prone to service level agreements (SLA) violation under high load and time-varying channel conditions, implying that an SLA enforcement system is needed in ITU-T's 5 G ML framework.

Modern mobile handheld devices, such as smartphones and tablets, feature multiple wireless interfaces, some of which can support device-to-device communications, which enable mesh networks on even when the infrastructure is unavailable. One of the key technological challenges hampering the use of multi-hop mesh networks is the extremely high communication overhead of route discovery and maintenance algorithms. The problem is especially pronounced under dynamic network conditions caused by user mobility and nodes joining and leaving the network. In this paper, we propose a fully distributed algorithm for constructing a virtual coordinate system used for geo-like routing by approximating the physical network nodes coordinate. The proposed algorithm, called gradient assisted routing (GAR), builds upon two-hop neighbors’ information exchanged in beacons in contrast to conventional geographic routing protocols which rely on external positioning information. We evaluate the proposed solution using algorithmic, topological, and routing-related metrics of interest. We further numerically quantify how the node mobility increases the time needed for topology stabilization, and how network size affects the route discovery success rate. Our comparison also shows that for small to mid-size mesh networks (up to 60 nodes), the performance of the proposed routing procedure is similar to the conventional geographic routing protocols that exploit external positioning information. The proposed solution may efficiently supplement the traditional on-demand routing in small to mid-size mesh systems by independently establishing 50 to 70% of paths and thereby reducing the discovery overheads.