The rapid expansion of the Internet of Things(IoT) has underscored the critical need for efficient and autonomous communication systems to sustain the massive, interconnected network of smart devices. Central to this challenge is optimizing communication protocols to ensure energy efficiency, reliability, and self-configurability across diverse IoT applications. This paper reviews the essence of leveraging artificial intelligence, specifically deep reinforcement learning, distributed AI services, swarm intelligence, metaheuristic optimization, and cross-layer approaches, for autonomous optimization and configuration in IoT and IoE (Internet of Everything) environments within the context of emerging 6 G technologies. It also implies a focus on comparing various methodologies and approaches to achieve efficient communication systems for IoT. The key criteria used in this comparison study are energy efficiency, transmission power, protocols, scalability, and Quality of service (QoS). By synthesizing these findings, our study highlights the strengths, limitations, and potential synergies between different approaches, offering insights into the future direction of IoT communication optimization.

The Internet of Things (IoT) continues to grow at remarkable speed, connecting different devices and facilitating communication and data exchange. Ambient Backscatter Communication (AmBC) has emerged as a low-power, low-cost alternative suitable for the connectivity of IoT. However, this technology still requires development to be adopted on a large scale. In this paper, we present a brief overview of AmBC, and highlight how it is appealing for IoT. We then go over the major challenges that AmBC faces with a concise explanation for each, as well as outline the directions future research should take to enhance the performance AmBC and push its integration with IoT forward. We also include a survey of some contributions in this domain.

Federated Learning, often known as FL, is an approach that has recently emerged as a potentially helpful method for training machine learning models in a distributed manner without the requirement of central data storage. However, when attempting to aggregate information, the inherent variety and discrepancies in the data contributed by many FL contributors might be a substantial obstacle. In order to address this problem, researchers have offered various solutions, one of which is called knowledge distillation (KD). Such a solution seeks to transfer knowledge from a larger, more precise model to a smaller model, thus enhancing its performance. This study provides a detailed examination of the effectiveness of KD in responding to these challenges posed by FL. We comprehensively review existing research, emphasizing the benefits and limitations of using these techniques in FL and discussing the numerous challenges and research questions in this field.

The rapid growth of Internet of Things (IoT) devices has increased the demand on cellular networks, particularly within Narrowband Internet of Things (NB-IoT) technology. To address the challenge of limited cellular capacity, this paper proposes an optimization method that groups devices and allocates specific time slots for each group to engage with the base station. The aim is to minimize interference and packet collisions, ensuring efficient device connectivity. We explore various clustering algorithms, including K-means, uniform repetition, and genetic algorithms (GA), with a focus on GA for its superior performance in reducing interference and improving connection success rates. Simulations confirm that the GA-based approach effectively manages device groups, enhances connectivity, and optimizes cell capacity in NB-IoT networks.

Localization of transmitters has been gaining interest increasingly. Due to the emergence of directional transmitters in future technologies, localization in this manner is particularly considered. However, limited studies consider the localization of multiple directional transmitters. Our proposed system model describes multiple directional emitters by the means of a Uniform Linear Array (ULA). To localize these transmitters, we adapt the classical Direction of Arrival (DoA) localization techniques with our model. The localization accuracy is studied for multiple affecting parameters and compared to the theoretical limit of Cramer Rao Bound (CRB). The results show that a directional transmitter can be localized by the means of its high sidelobe effects. Root MUSIC outperforms other techniques in localizing the users.

Intelligent reflecting surfaces have been recently introduced as a promising technology to achieve smart and reconfigurable radio environment. Unlike traditional approaches that consider the environment to be uncontrollable in wireless networks, intelligent reflecting surfaces are capable of smartly modifying the wireless channel. This opens a new paradigm with a lot of expected gains that need to be identified and investigated. In this paper, we exploit an intelligent reflecting surface module to reduce the power demand of a base station that is serving multiple users. The problem of managing the intelligent surface's elements is formulated as a non-linear integer problem. Due to the complexity of the problem, the intelligent reflecting surface is discretized into blocks of elements. The blocks have different sizes to assist users with different conditions. We propose a two-stage approach that selects the users to be assisted and then allocates the blocks of elements to the selected users. We also compare the block allocation to the optimal elements allocation for selected users to show the effectiveness of the proposed approach. Results show that the power need of a base station can be decreased by more than 8% using an intelligent reflecting surface module with limited number of elements if the module is well-positioned and assisted users are properly selected.

We consider a cognitive radio network with a primary user (PU) and secondary users (SU), all equipped with multiple antennas to exploit the spatial characteristics for transmission. The SUs cooperate to estimate the beam of the PU's signal. Thus, the available space is divided into two parts. The first is occupied by the PU and should be restricted on the SUs, whereas the second is highly accessible by the SUs transmissions. The beam-estimation accuracy is studied based on two metrics, namely, the angular missed and false detection. The results showed that the accuracy increases for a Rician channel rather than a Rayleigh fading channel. Additionally, no matter how far the SUs' range of distribution is, a high number of SUs can accurately estimate the beam. A zero angular missed detection occurs for many SUs, with a tax of slightly increasing the angular false detection.

In Wireless Sensor Networks (WSN), the nodes may be randomly deployed over a harsh geographical zone. Usually, these nodes are battery powered with limited transmission and processing capabilities. Managing the residual energy is very crucial in such network, since replacing a battery may not be always feasible. Energy is essentially consumed during packet transmission phase. Therefore, routing protocols become of a high importance since they impact on energy consumption during data transmissions. In this paper, we present a new routing protocol based on the geographical location information in an uplink multi-hop WSN. The proposed protocol assumes that the geographical zone and the transmitted data are more precious than the sender node authentication. Our proposed protocol guarantees the delivery of data packet in a dense network without any knowledge of the topology nor the path nodes between the source and the destination. We analytically derive the average consumed energy, the probability of packet loss, and the mean delivery time. Numerical results corroborate the superiority of our protocol over the state-of-the-art Distance Routing Protocol (DIR) in terms of connectivity, lifespan, memory, and latency.

Satisfying the users’ increasing demand and reducing the networks’ energy consumption are among the most critical requirements of future cellular networks. In this paper, we exploit Intelligent Reflecting Surfaces (IRSs) to reduce the bandwidth required by users, which will allow more users to be served and/or reduce the energy footprint of cellular base stations. In contrast to most of the existing studies that focus on configuring the phase shifts of IRSs and/or the active beamforming of the base station, we consider that the IRS consists of blocks of resources that can be shared by several users. We formulate the problem of managing these resources as nonlinear integer problem. Then, we solve the optimization problem using exhaustive search, and propose two low complexity heuristic algorithms. The performance of the system is evaluated considering variable number of users, position of IRS, required bit rate and radius of the cell. Results show that using IRS can achieve significant bandwidth savings and important energy demand reduction when the IRS resources are well managed.

Wireless sensor networks are becoming very popular nowadays. Sensors in such networks are used to gather data periodically about a given zone area and send the collected data to the sink. However, such networks face several challenges especially the limited energy source and the data management. Hence, data sampling approach is becoming one of the essential techniques that saves he sensor energies and extend the network lifetime. Adaptive algorithms are created to allow each sensor to adapt its sampling rate to the application under surveillance, which leads to reduced data collection thus, reducing energy consumption. In this paper, we propose a new adaptive sampling technique that is dedicated to periodic sensor network applications. Our technique consists of two stages: aggregation and adapting stages. The first stage is applied at sensor level and aims to reduce the amount of data collected by the sensor. The second stage is applied at an intermediate level node called cluster-head (CH). The CH receives data periodically from the sensors and computes the new sampling rate for each sensor based on the spatio-temporal correlation between the sensors. Our technique is evaluated based on real sensor data collected in the Intel lab. The obtained results show the effectiveness of our technique in terms reducing the energy consumption while ensuring a high level of data accuracy and coverage network.

In this paper, we present a new Cognitive Radio (CR) paradigm based on the Full-Duplex mechanism. In the classical Full-Duplex CR (FD-CR) system, the Secondary User (SU) can examine the availability of a channel while transmitting. This fact leads to enhance the SU transmission rate. However, in this classical secondary network, SU is assumed to adopt frequency or time division duplex. In the proposed work, we analyse the CR performance with an In-Band Full-Duplex communication, i.e. SU can simultaneously receive and transmit at the same band. This scenario is accompanied with a Full-Duplex sensing, i.e. SU performs the Spectrum Sensing while transmitting. The Spectrum Sensing performance is analysed as well as proposing an adaptable detection mechanism that can perform well under such situation. Further, a study on the SU throughputs is developed in order to show the spectral efficiency of the proposed CR paradigm.