Due to their adaptability and mobility, Unmanned Aerial Vehicles (UAVs) are becoming increasingly essential for wireless network services, particularly for data harvesting tasks. In this context, Artificial Intelligence (AI)-based approaches have gained significant attention for addressing UAV path planning tasks in large and complex environments, bridging the gap with real-world deployments. However, many existing algorithms suffer from limited training diversity and limited generalization beyond the training distribution, which hampers their performance in highly dynamic environments. Moreover, they often overlook the inherently multi-objective nature of the task, treating it in an overly simplistic manner. To address these limitations, we propose an attention-based Multi-Objective Reinforcement Learning (MORL) architecture that explicitly handles the trade-off between data collection and energy consumption in urban environments, even without prior knowledge of wireless channel conditions. Our method learns a single model capable of adapting to varying trade-off preferences and dynamic scenario parameters without the need for fine-tuning or retraining. Extensive simulations show that our approach achieves substantial improvements in performance, model compactness, sample efficiency, and most importantly, generalization to previously unseen scenarios, outperforming existing RL solutions.

This paper investigates the performance of cellular networks assisted by unmanned aerial vehicles (UAVs) acting as flying base stations (FlyBSs). We focus on a scenario with multi-hop relaying via FlyBSs to deliver data from a ground base station (GBS) to users in a challenging case with the channels reused at all hops to exploit radio resources efficiently. Our objective is to maximize the sum capacity of the users via an optimization of FlyBSs’ position in 3D, association of users to either GBS or to one of the FlyBSs, allocation of channels for communication at individual hops, and allocation of transmission power for all channels. Moreover, practical constraints on the FlyBSs’ movement, transmission and propulsion power, and backhaul capacity are taken into account. Due to a non-convexity and discreetness of the objective and some constraints, there is no optimal solution to the formulated problem. Thus, we propose an analytical solution based on an alternating optimization of an energy-efficient placement of the FlyBSs, channel allocation, user association, and transmission power. Each subproblem in the alternating optimization is substituted either by a linear programming (LP) problem through a change of variables, or by a convex problem via a conversion of the objective and constraints. The results show an increase in sum capacity by 35%–60% compared to related works while the FlyBSs’ propulsion power consumption is not increased.

In mobile networks, unmanned aerial vehicles (UAVs) acting as flying base stations (FlyBSs) can effectively improve performance. Nevertheless, such potential improvement requires an efficient positioning of the FlyBS. In this paper, we study the problem of sum downlink capacity maximization in FlyBS-assisted networks with mobile users and with a consideration of wireless backhaul with channel reuse while a minimum required capacity to every user is guaranteed. The problem is formulated under constraints on the FlyBS's flying speed, propulsion power consumption, and transmission power for both of flying and ground base stations. None of the existing solutions maximizing the sum capacity can be applied due to the combination of these practical constraints. This paper pioneers in an inclusion of all these constraints together with backhaul to derive the optimal 3D positions of the FlyBS and to optimize the transmission power allocation for the channels at both backhaul and access links as the users move over time. The proposed solution is geometrical based, and it shows via simulations a significant increase in the sum capacity (up by 19%-47%) compared with baseline schemes where one or more of the aspects of backhaul communication, transmission power allocation, and FlyBS's positioning are not taken into account.

In this work, we investigate the problem of an online trajectory design for an Unmanned Aerial Vehicle (UAV) in a Federated Learning (FL) setting where several communities exist, each defined by a unique task to be learned. In this setting, spatially distributed devices belonging to each community collaboratively contribute towards training their community model via wireless links provided by the UAV. Accordingly, the UAV acts as a mobile orchestrator coordinating the transmissions and the learning schedule among the devices in each community, intending to accelerate the learning process of all tasks. We propose a heuristic metric as a proxy for the training performance of the different tasks. Capitalizing on this metric, a surrogate objective is defined which enables us to jointly optimize the UAV trajectory and the scheduling of the devices by employing convex optimization techniques and graph theory. The simulations illustrate the out-performance of our solution when compared to other handpicked static and mobile UAV deployment baselines.

Deployment of multi-hop network of unmanned aerial vehicles (UAVs) acting as flying base stations (FlyBSs) presents a remarkable potential to effectively enhance the performance of wireless networks. Such potential enhancement, however, relies on an efficient positioning of the FlyBSs as well as a management of resources. In this paper, we study the problem of sum capacity maximization in an extended model for mobile networks where multiple FlyBSs are deployed between the ground base station and the users. Due to an inclusion of multiple hops, the existing solutions for two-hop networks cannot be applied due to the incurred backhaul constraints for each hop. To this end, we propose an analytical approach based on an alternating optimization of the FlyBSs' 3D positions as well as the association of the users to the FlyBSs over time. The proposed optimization is provided under practical constraints on the FlyBS's flying speed and altitude as well as the constraints on the achievable capacity at the backhaul link. The proposed solution is of a low complexity and extends the sum capacity by 23%-38% comparing to state-of-the-art solutions.

The use of unmanned aerial vehicles (UAVs) acting as flying base stations (FlyBSs) is considered as an effective tool to improve performance of the mobile networks. Nevertheless, such potential improvement requires an efficient positioning of the FlyBS. In this paper, we maximize the sum downlink capacity of the mobile Internet of Things devices (IoTD) served by the FlyBSs while a minimum required capacity to every device is guaranteed. To this end, we propose a geometrical approach allowing to derive the 3D positions of the FlyBS over time as the IoTDs move and we determine the transmission power allocation for the IoTDs. The problem is formulated and solved under practical constraints on the FlyBS's transmission and propulsion power consumption as well as on flying speed. The proposed solution is of a low complexity and increases the sum capacity by 15% -46% comparing to state-of-the-art works.

Deep Reinforcement Learning (DRL) is gaining attention as a potential approach to design trajectories for autonomous unmanned aerial vehicles (UAV) used as flying access points in the context of cellular or Internet of Things (IoT) connectivity. DRL solutions offer the advantage of on-the-go learning hence relying on very little prior contextual information. A corresponding drawback however lies in the need for many learning episodes which severely restricts the applicability of such approach in real-world time- and energy-constrained missions. Here, we propose a model-aided deep Q-learning approach that, in contrast to previous work, considerably reduces the need for extensive training data samples, while still achieving the overarching goal of DRL, i.e to guide a battery-limited UAV on an efficient data harvesting trajectory, without prior knowledge of wireless channel characteristics and limited knowledge of wireless node locations. The key idea consists in using a small subset of nodes as anchors (i.e. with known location) and learning a model of the propagation environment while implicitly estimating the positions of regular nodes. Interaction with the model allows us to train a deep Q-network (DQN) to approximate the optimal UAV control policy. We show that in comparison with standard DRL approaches, the proposed model-aided approach requires at least one order of magnitude less training data samples to reach identical data collection performance, hence offering a first step towards making DRL a viable solution to the problem.

In this work, we develop an algorithm to construct radio maps that can predict the received signal strength between a UAV-mounted base station and arbitrary ground users. The novelty of the work lies in the fact that these maps are constructed by fusing UAV-user radio signal strength measurements, and depth information of the surrounding environment which is obtained by an on-board laser range finder sensor. The proposed approach exploits both line-of-sight (LoS) and non-line-of-sight (NLoS) nature of UAV-user channels and depth information to first obtain the 3D map of the city and then later use it to estimate the radio map. Numerical results demonstrate the significant gain brought by the fusion of radio and depth measurements as opposed to a system which only relies on radio measurements.

This article considers the problem of localizing outdoor ground radio users with the help of an unmanned aerial vehicle (UAV) on the basis of received signal strength (RSS) measurements in an urban environment. We assume that the propagation model parameters are not known a priori, and depending on the UAV location, the UAV-user link can experience either Line-of-Sight (LoS) or Non-Line-of-Sight (NLoS) propagation condition. We assume that a 3-D map of the environment is available which the UAV can exploit in the localization process. Based on the proposed map-aided estimator, we devise an optimal UAV trajectory to accelerate the learning process under a limited mission time. To do so, we borrow tools, such as Fisher information from the theory of optimal experiment design. Our map-aided estimator achieves superior localization accuracy compared to the map-unaware methods, and our simulations show that optimized UAV trajectory achieves superior learning performance compared to random trajectories.

Cellular connected unmanned aerial vehicles (UAVs) that can operate safely in beyond visual line of sight conditions are expected to open important future opportunities in the areas of transportation, goods delivery, and system monitoring. A key challenge in this area lies in the design of trajectories which, while allowing the completion of the UAV mission, can guarantee reliable cellular connectivity all along the path. Previous approaches in this domain have considered either simplistic propagation model assumptions (e.g. Line of Sight based) or more advanced models but with computationally demanding optimization solutions. In this paper, we propose a novel approach for trajectory design using a coverage map that can be obtained with a combination of a 3D map of the environment and radio propagation models. Leveraging on the convexity of subregions within the coverage map, we propose a low-complexity graph based algorithm which is shown to achieve quasi-optimal performance at a fraction of the computational cost of known optimal methods.

We consider a scenario where an unmanned aerial vehicle (UAV)-mounted flying base station is providing data communication services to a number of radio nodes spread over the ground. We focus on the problem of resource-constrained UAV trajectory design with: 1) optimal channel parameters learning and 2) optimal data throughput as key objectives, respectively. While the problem of throughput optimized trajectories has been addressed in prior works, the formulation of an optimized trajectory to efficiently discover the propagation parameters has not yet been addressed. When it comes to the communication phase, the advantage of this paper comes from the exploitation of a 3-D city map. Unfortunately, the communication trajectory design based on the raw map data leads to an intractable optimization problem. To solve this issue, we introduce a map compression method that allows us to tackle the problem with standard optimization tools. The trajectory optimization is then combined with a node scheduling algorithm. The advantages of the learning-optimized trajectory and of the map compression method are illustrated in the context of intelligent Internet of Things data harvesting.

This paper considers the problem of 3D city map reconstruction. The key novelty here lies in the sole exploitation of UAV-bound radio measurements as a way to recover the map data, i.e. no image of the city is taken or processed. The proposed approach relies on the unique ability for a UAV- to-ground communication system to detect and classify line-of-sight (LoS) vs. non line-of-sight (NLoS) channels towards ground users using machine learning tools. Once classification is carried out, the LoS vs. NLoS data is fed as input to a building position and height reconstruction algorithm. The map reconstruction quality is analyzed as a function of user density and UAV altitude, revealing the notion of an optimal height for the UAV which is predicted using an analytical model.