Path planning and optimization for unmanned aerial vehicles (UAVs)-assisted next-generation wireless networks is critical for mobility management and ensuring UAV safety and ubiquitous connectivity, especially in dense urban environments with street canyons and tall buildings. Traditional statistical and model-based techniques have been successfully used for path optimization in communication networks. However, when dynamic channel propagation characteristics such as line-of-sight (LOS), interference, handover, and signal-to-interference and noise ratio (SINR) are included in path optimization, statistical and model-based path planning solutions become obsolete since they cannot adapt to the dynamic and time-varying wireless channels, especially in the mmWave bands. In this paper, we propose a novel model-free actor-critic deep reinforcement learning (AC-DRL) framework for path optimization in UAV-assisted 5G mmWave wireless networks, which combines four important aspects of UAV communication: flight time, handover, connectivity and SINR. We train an AC-RL agent that enables a UAV connected to a gNB to determine the optimal path to a desired destination in the shortest possible time with minimal gNB handover, while maintaining connectivity and the highest possible SINR. We train our model with data from a powerful ray tracing tool called Wireless InSite, which uses 3D images of the propagation environment and provides data that closely resembles the real propagation environment. The simulation results show that our system has superior performance in tracking high SINR compared to other selected RL algorithms.

With the advent of Advanced Driver Assistance Systems (ADAS) and intelligent transport system applications, recognizing driver emotions has become essential for a decision support system (DSS) with humans in the loop (HITL). Multimodal approaches using visual cues, speech, physiological signals, and driving patterns improve emotion recognition but are challenging in resource-constrained environments where only a subset of modalities is available. This work addresses these challenges by combining multi-modal benefits with single-modality inference for emotion recognition using unlabeled external road condition data. Unlike traditional methods that average teachers' contribution, the proposed cross-modal distillation (CMD) weights teachers thanks to the Shapley additive global explanation (SAGE) aid, which improves the student model's accuracy and provides an interpretation of it. Experimental evaluations of the PPBEmo dataset show that XA-CMD improves emotion recognition accuracy with other baselines and provides deeper insights into decision-making.

Modern vehicles are equipped with sophisticated systems that continuously monitor both their mechanical condition and the well-being of passengers. In the vehicular network scenario, this availability of a vast amount of data has encouraged ever more the development of ML-based systems to enable highly reliable functionalities for supporting autonomous driving. However, a major challenge is to enable vehicles to process ML-based complex models quickly and efficiently. This problem can be solved by utilizing edge-based computing solutions where computing and storage resources available in network infrastructures are located near the vehicles. By offloading some of the processing tasks to these local resources, vehicles can achieve faster response times and enhanced efficiency. In this paper, we analyze the computing and communication performance of a federated multimodal distillation approach for driver emotion detection for a vehicular communication semi-urban scenario, which will be modeled by implementing stochastic geometry models. The aim of our analysis is to evaluate the impact of a complex ML-based approach on the communication infrastructure, where locally distilled models suitable for constrained devices are federated into a global model. We also investigate the impact on the network of the load due to the learning procedure when IID data and non-IID are considered. The numerical results highlight the strengths and weaknesses of the communication infrastructure when heterogeneous wireless technologies such as 5G and WiFi are involved in handling this type of approach for vehicles with limited computational resources.

The emergence of 6G networks demands environment-aware communication paradigms to ensure reliable and efficient connectivity, and Channel Knowledge Maps (CKMs) offer a promising solution by mapping spatial locations to detailed channel characteristics for proactive network optimization. In this context, this paper proposes an explainable Machine Learning (ML)-based framework that uses geometrical features to predict receiver state probabilities in UAV-based mmWave communication networks. Geometrical characteristics extracted from the environment surrounding each receiver are used to train ML models, namely Decision Tree (DT), K-Nearest Neighbors (KNN), and Deep Neural Network (DNN) models, to predict three receiver states probabilities: Line-of-Sight (LOS), No-Line-of-Sight (NLOS), and Blocked. Experimental results show that the DNN model outperforms DT and KNN, achieving higher accuracy across all states, albeit with no inherent explainability. To address this, the SHapley Additive exPlanations (SHAP) method is applied to indicate feature contributions to each state prediction of the black-box DNN model. This improves the interpretability and reliability of the proposed environment-aware framework for$\mathbf{6 G}$UAV-based networks.

Emotion recognition has attracted a lot of interest in recent years in various application areas such as healthcare and autonomous driving. Existing approaches to emotion recognition are based on visual, speech, or psychophysiological signals. However, recent studies are looking at multimodal techniques that combine different modalities for emotion recognition. In this work, we address the problem of recognizing the user’s emotion as a driver from unlabeled videos using multimodal techniques. We propose a collaborative training method based on cross-modal distillation, i.e., “FedCMD” (Federated Cross-Modal Distillation). Federated Learning (FL) is an emerging collaborative decentralized learning technique that allows each participant to train their model locally to build a better generalized global model without sharing their data. The main advantage of FL is that only local data is used for training, thus maintaining privacy and providing a secure and efficient emotion recognition system. The local model in FL is trained for each vehicle device with unlabeled video data by using sensor data as a proxy. Specifically, for each local model, we show how driver emotional annotations can be transferred from the sensor domain to the visual domain by using cross-modal distillation. The key idea is based on the observation that a driver’s emotional state indicated by a sensor correlates with facial expressions shown in videos. The proposed “FedCMD” approach is tested on the multimodal dataset “BioVid Emo DB” and achieves state-of-the-art performance. Experimental results show that our approach is robust to non-identically distributed data, achieving 96.67% and 90.83% accuracy in classifying five different emotions with IID (independently and identically distributed) and non-IID data, respectively. Moreover, our model is much more robust to overfitting, resulting in better generalization than the other existing methods.

In Non-Terrestrial Networks (NTNs), where LEO satellites and User Equipment (UE) move relative to each other, Line-of-Sight (LOS) tracking, and adapting to channel state variations due to endpoint movements are a major challenge. Therefore, continuous LOS estimation and channel impairment compensation are crucial for a UE to access a satellite and maintain connectivity. In this paper, we propose a Actor-Critic (AC)-Reinforcement Learning (RL) framework for traffic scheduling in NTN scenarios where the channel state is non-stationary due to the variability of LOS, which depends on the current satellite elevation. We deploy the framework as an agent in a Multi-Path Routing (MPR) scheme where the UE can access more than one satellite simultaneously to improve link reliability and throughput. We study how the agent schedules traffic on multiple satellite links by adopting the AC version of RL. The agent continuously trains based on variations in satellite elevation angles, handoffs, and relative LOS probabilities. We compare the agent retraining time with the satellite visibility intervals to investigate the effectiveness of the agent's learning rate. We carry out performance analysis considering the dense urban area of Chicago, where high-rise buildings significantly affect the LOS. The simulation results show how the learning agent selects the scheduling policy when it is connected to a pair of satellites. The results also show that the retraining time of the learning agent is up to 0.1 times the satellite visibility time at certain elevations, which guarantees efficient use of satellite visibility.

The paper proposes a data-driven approach to air-to-ground channel estimation in a millimeter-wave wireless network on an unmanned aerial vehicle. Unlike traditional centralized learning methods that are specific to certain geographical areas and inappropriate for others, we propose a generalized model that uses Federated Learning (FL) for channel estimation and can predict the air-to-ground path loss between a low-altitude platform and a terrestrial terminal. To this end, our proposed FL-based Generative Adversarial Network (FL-GAN) is designed to function as a generative data model that can learn different types of data distributions and generate realistic patterns from the same distributions without requiring prior data analysis before the training phase. To evaluate the effectiveness of the proposed model, we evaluate its performance using Kullback-Leibler divergence (KL), and Wasserstein distance between the synthetic data distribution generated by the model and the actual data distribution. We also compare the proposed technique with other generative models, such as FL-Variational Autoencoder (FL-VAE) and stand-alone VAE and GAN models. The results of the study show that the synthetic data generated by FL-GAN has the highest similarity in distribution with the real data. This shows the effectiveness of the proposed approach in generating data-driven channel models that can be used in different regions.

A distributed decision support system has been developed to assist seafarers during several navigation tasks, for instance, in avoiding a collision with a detected obstacle in the sea and envisioning a future autonomous navigation system. In this paper, the decision support system is based on the results of a customized simulation model representing the ship’s behavior, including hydrodynamics, propulsion, and control effects. Sensors monitor and collect the parameters of the environment and the ship onboard. The telemetry and the calculated route are visualized on a wearable visor exploiting augmented reality. Such context information is also replicated ashore through a narrow-band satellite link using an IoT publish-subscribe communication paradigm to allow one or more remote seafarers to supervise the situation in a virtual reality environment. Overall, the potential of the proposed system is presented and discussed for application in the context of autonomous navigation.

Automatic traffic classification is increasingly important in networking due to the current trend of encrypting transport information (e.g., behind HTTP encrypted tunnels) which prevent intermediate nodes to access end-to-end transport headers. This paper proposes an architecture for supporting Quality of Service (QoS) in hybrid terrestrial and SATCOM networks based on automated traffic classification. Traffic profiles are constructed by machine-learning (ML) algorithms using the series of packet sizes and arrival times of QUIC connections. Thus, the proposed QoS method does not require explicit setup of a path (i.e. it provides soft QoS), but employs agents within the network to verify that flows conform to a given traffic profile. Results over a range of ML models encourage integrating ML technology in SATCOM equipment. The availability of higher computation power at low-cost creates the fertile ground for implementation of these techniques.

Reinforcement Learning (RL) has recently found wide applications in network traffic management and control because some of its variants do not require prior knowledge of network models. In this paper, we present a novel scheduler for real-time multimedia delivery in multipath systems based on an Actor-Critic (AC) RL algorithm. We focus on a challenging scenario of real-time video streaming from an Unmanned Aerial Vehicle (UAV) using multiple wireless paths. The scheduler acting as an RL agent learns in real-time the optimal policy for path selection, path rate allocation and redundancy estimation for flow protection. The scheduler, implemented as a module of the GStreamer framework, can be used in real or simulated settings. The simulation results show that our scheduler can target a very low loss rate at the receiver by dynamically adapting in real-time the scheduling policy to the path conditions without performing training or relying on prior knowledge of network channel models.

Providing reliable and efficient connectivity for maritime systems is a key objective to enable new services and to offer anytime-anywhere communication solutions to vessels operating in remote areas (e.g. oceans). Such a task has recently become quite compelling because of the increasing maritime traffic, which should be supported by an information distribution infrastructure in order to guarantee efficient data communication among vessels and control centres on land. To this end, satellites are the perfect candidates for achieving such an objective. This paper focuses on the case of messaging from moving vessels to fixed control centres on land by means of advanced channel random access schemes exploiting time diversity. In more detail, the paper extends an existing theoretical framework to evaluate the packet loss probability in Framed Slotted ALOHA systems by delving into the probability distribution of both colliding users within a frame and their replicas in time slots. The theoretical framework is validated through simulations campaigns, showing a good match between analytical and simulated results.

This work describes a real testbed for enabling Unmanned Aerial Vehicles (UAVs)-to-ground real-time video streaming. The aim is in providing a video feed to the pilot on the ground for Beyond Visual Line of Sight (BVLoS) operations exploiting cellular networks in urban/suburban areas. Towards this aim, multipath communications are used in a multi-operator setup to counteract the intermittent network coverage in urban and suburban areas. The main requirements are low latency and a continuous video stream of reasonable quality. We rely on both upper-layer Forward Error Correction (FEC) techniques and link diversity, so to increase the probability of fluid video playback with acceptable quality. We report per-link statistics, collected during field trials, of three different cellular operators, to analyse the impact of using a set of physical links as a single logical one on an RTP-based video streaming. In our tests, such a setup has provided a good level of performance.

Nowadays, Machine-to-Machine and Internet of Things traffic sources puts the terrestrial networks under great pressure. While 5G is still on its way, satellites are used to deliver a fraction of such an enormous traffic rate. In this work, we investigate the use of the Constrained Application Protocol (CoAP) to reliably deliver Machine-to-Machine and Internet of Things traffic in a push fashion, which also implements a Selective Repeat Automatic Repeat reQuest and a sender-based variant of the TCP Friendly Rate Control protocol. We aim at providing an analytical model to evaluate the working point of the system in non-saturated conditions as a function of the MAC parameters in use, when such a closed-loop congestion control mechanism is in use over a random access satellite channel. The proposed analytical model is then validated against simulation results, showing a good precision.

Device-free localization algorithms attract, among others, the attention of researchers working in the Ambient Assisted Living (AAL) scenarios, where the target user might not be able or willing to wear any devices. We concentrate on systems that exploit the Received Signal Strength indicator coming from wireless devices whose position is known, called anchors. In this paper we select and test the main device-free localization solutions and experimentally compare their performance using a smaller number of anchors than commonly found in the literature. We illustrate the procedure used to validate our comparing procedure and we give suggestions on usability in the application scenarios typical of AAL. To the best of our knowledge, this is the first direct comparison between different device-free algorithms using the same input data for all of them, and the first one that compares their performance with a varying number of anchors. Thanks to the characteristics of our comparison procedure, we can make suggestions about the more appropriate algorithms to use for different kinds of applications.

Indoor localization applications that involve Wireless Sensor Networks (WSNs) identify the target position by measuring the Received Signal Strength (RSS), the Time of Arrival (ToA), the Time Difference of Arrival (TDoA) or the Angle of Arrival (AoA). Of these, the most promising for low-cost applications are those based on measures of the RSS, which exploit the relationship between RSS and the distance, or more reliably the relation between the multi-path interference (shadowing) and the position of the target. These methods work with WSNs based on Wi-Fi, Bluetooth and ZigBee sensor technologies. In this paper we concentrate on device-free RSS-based indoor localization methods. These methods, which have generated much research interest in the last few years, are now starting to hit the market. Specifically, the purpose of this paper is to assess the performance improvements of a Variance-based Radio Tomographic Imaging technique, when scanning various radio channels with respect to using only one, the latter being the “minimum introduced interference” option. Moreover, in this paper we will discuss in which application scenario the multi-channel scanning technique is usable and appropriate. The experimental data used for target localization are captured by wireless sensors deployed in the localization area and the localization error metrics include the mean square error and percentiles of the error distribution. Specifically, we aim to study the localization error reduction obtained by using multiple ZigBee channels, with respect to using a single channel.