The wide adoption of large-scale Internet of Things (IoT) systems has led to an unprecedented increase in backhaul (BH) traffic congestion, making it critical to optimize traffic management at the network edge. In IoT systems, the BH network is supported by various backhauling technologies that have different characteristics. Also, the characteristics of the BH links can be sometimes time varying and have an unknown state, due to external factors such as having the resources shared with other systems. It is the responsibility of the edge devices to be able to forward IoT traffic through the unknown-state BH network by selecting the suitable BH link for each collected data flow. To the best of our knowledge, this type of BH selection problem is not addressed in the literature. Therefore, there is a crucial need to develop intelligent approaches enabling edge devices to learn how to deal with unknown-state (partially observable) components of the BH network, which is the primary goal of this article. We propose an edge-based BH selection technique for improving traffic delivery by exploiting multiobjective feedback on delivery performance. The proposed approach relies on the advantage-actor–critic deep reinforcement learning (DRL) methods. Moreover, to improve the DRL training performance in large-scale deployments of distributed IoT systems, federated learning (FL) is applied to enable multiple edge devices to collaborate in training a shared BH selection policy. The proposed federated DRL (F-DRL) approach is able to solve the BH selection problem as verified and demonstrated through extensive simulations.

Internet of Things (IoT) systems are driven by the massive data generation at the sensing layer. Most of data management protocols in the sensing layer target optimizing different Quality-of-Service (QoS) metrics, such as data rate, packet loss, and delay, while taking the limitations of wireless sensing networks (WSNs) into consideration. However, it is also critical to consider improving the quality of collected data, especially when the WSNs are congested by delay-sensitive data packets, leading to high packet loss. While Value-of-Information (VoI), which represents data freshness and time relevance, is commonly used to define data value, it does not provide a notion of data recoverability and tolerance-to-loss. In this article, the main contribution is to develop data management schemes to cope with inevitable data loss by identifying the data portion with a higher value to the overlaying applications. We also develop a novel metric that is referred to as information content (IC), quantifying the amount of information in data. The IC is defined such that data holding information of low-probable events have higher IC than data holding information of high-probable events. In this context, the VoI and IC are exploited in developing information-oriented traffic forwarding and reduction schemes to ensure that all dropped packets are more accurately recoverable through a traffic recovery scheme, and therefore the running applications are not disrupted. Through extensive Monte-Carlo simulations, we show that the proposed information-oriented data management improves the performance in terms of data congestion, lifetime, packet loss, delay, and data recovery accuracy.

The next generation of Mission Critical Networks (MCNs) will likely be based on Long Term Evolution (LTE) technology. This is due to the many features LTE can offer such as reliable broadband communications and Proximity Services (ProSe). However, the environments where MCNs mostly work require irregular deployment of LTE. Such requirements can impose constraints on the Frequency Reuse (FR) algorithms and Medium Access Control (MAC) schedulers to be used. In this paper, we develop a high-level parameter selection framework to enable the testing of irregular deployments of LTE-based MCNs. The developed tool provides a Graphical User Interface (GUI) that can be easily used to change the system settings and the available resources in order to find the most suitable system parameters. The developed algorithm also considers minimizing the used spectrum and allows the end-user to select a Quality-of-Service (QoS) threshold to control the system performance. Results show that the proposed framework provides accurate recommendations on the needed resources, FR algorithms, and MAC schedulers to be used. Insights on the system performance and suggested deployment style are also provided.

Spatial and temporal correlation among the generated traffic in wireless sensor networks (WSNs) can be exploited in reducing the energy consumption of continuous sensor data collection. Dual prediction (DP) and data compression (DC) schemes rely on the spatio-temporal correlation to reduce the number of transmissions across WSNs, which leads to conserving energy and bandwidth. In this paper, we present both schemes in a two-tier data reduction framework. The DP scheme is used to reduce transmissions between cluster nodes and cluster heads, while the DC scheme is used to reduce traffic between cluster heads and sink nodes. For both schemes, various algorithms will be studied and compared in terms of accuracy, delay, and transmission reduction percentage. For the DP scheme, neural networks (NNs) and long short-term memory networks (LSTMs) are proposed to perform predictions. The training phase of the NNs and LSTMs is done online which is necessary in the DP scheme. The performance will be compared to popular least-mean-square approaches. Regarding the DC scheme, principal component analysis, non-negative matrix factorization, truncated-singular value decomposition, and discrete wavelet transform will be discussed and compared. This paper focuses on comparative analysis of various data reduction algorithms alongside the proposed ones. Finally, design challenges and open research areas for having more transmission reductions will be presented.

Due to the critical importance of Public Safety Networks (PSNs), they must have regular on-site performance evaluation. Conducting tests on real systems are expensive in terms of money, efforts, resources, and time. On the other hand, simulation and emulation tools are very important in the design and modeling of engineering systems. They can play a significant role in reducing the high expenses of such periodic tests. NS-3 is a widely used network simulator with a lot of capabilities. However, simulating realistic and large-scale PSNs reveals many limitations in the performance of simulation and urges for methods to speedup the simulations. Despite the exerted efforts in creating distributed simulators, less work that targets wireless networks is done. In this paper, we survey the different methods to enhance the simulation performance including parallel and distributed solutions in the PSN context. We also propose a framework to evaluate and test PSNs in real-time manner. This paper opens the door for future work in achieving better simulations, in terms of reliability and consumed time, for PSNs' large-scale networks.

Long Term Evolution (LTE) is a strong candidate for the next generation of Mission-Critical Networks (MCNs) for public safety and military use. However, extensive investigation on different operating scenarios has to be done before adopting LTE. Emulating LTE networks is the best solution to provide the needed insights to the regulators so that they can make informed decisions. In this paper, we integrate LTE-EPC Network simulator (LENA) module with the Common Open Research Emulator (CORE) in order to enable the emulation of large-scale LTE networks. The developed tool bridges the real and the simulated worlds by allowing us to connect real devices to simulated devices and passing traffic to the simulated world from the real world and vice versa. Using this tool, mission-critical scenarios could be investigated in a cost effective way.