Francesco Devoti, Placido Mursia, Vincenzo Sciancalepore, Xavier Costa-Pérez
©SHUTTERSTOCK.COM/AEROGONDO2
Aerial communication is gradually taking an assertive role within common societal behaviors by means of unmanned aerial vehicles (UAVs), high-altitude platforms (HAPs), and fixed-wing aircrafts (FWAs). Such devices can assist general operations in a diverse set of heterogeneous applications, such as video surveillance, remote delivery, and connectivity provisioning in crowded events and emergency scenarios. Given their increasingly higher technology penetration rate, telco operators started looking at the sky as a new potential direction to enable a three-dimensional (3D) communication paradigm. However, designing flying mobile stations involves addressing a daunting number of challenges, such as an excessive onboard control overhead, variable battery drain, and advanced antenna design. To this end, the newly born smart surfaces technology may come to help: reconfigurable intelligent surfaces (RISs) may be flexibly installed on board to control the terrestrial propagation environment from an elevated viewpoint by involving low-complex and battery-limited solutions. In this article, we focus on identifying the key optimization aspects to be considered when designing RIS-based aerial networks, and in particular the associated control architecture, by shedding light on novel use cases, corresponding requirements, and potential solutions.
Aerial communications, whereby flying devices such as UAVs, FWAs, or HAPs are used to facilitate the communication between two given nodes in the network, are a well-established technology capable of providing powerful yet flexible platforms to enhance current fifth-generation (5G) and beyond network infrastructures (6G) [1]. Such types of deployment, denoted as air-to-ground (A2G) networks, are envisioned as an integral part of future Internet of Things networks thanks to their agility to act as portable on-demand base stations (BSs), flexibility of movement, and an overall increased probability of high end-to-end channel quality. As a matter of fact, unlike fixed BSs, flying devices can be exploited on demand and reused for different applications, e.g., assisting rescue operations after natural disasters and (re)establishing a damaged or unavailable network infrastructure [2]. In addition, they can hover, avoiding obstacles that may cause blockage. Thus, they significantly increase the probability of establishing a line of sight (LoS) with terrestrial users, giving rise to the new 3D communication paradigm [3] in the 6G era.
However, despite the potential and envisioned business opportunities, aerial communications are hindered by the limited power budget and maximum carrying load of flying devices. For example, in the case of UAVs, it significantly constrains its feasibility: carrying bulky, heavy, and power-hungry equipment such as active antennas on board results in high design complexity and capital expenditure, further limiting the application range of aerial communications. Moreover, conventional A2G networks require signal processing capabilities on board, e.g., for channel estimation or precoding, and frequent backhaul communications, which further compound the design complexity of such systems.
To overcome the abovementioned issues, there has been growing interest in the emerging RIS technology, which is widely recognized as the means to greatly improve the quality of wireless communication links toward 6G networks (A2G communications) as depicted in Figure 1 [4]. RISs are 2D surfaces divided into unit cells spaced at a subwavelength distance, which can reflect the incoming signal by adding a given tunable phase shift. As a result, the reflected signal components can be combined constructively at the intended receiver position and destructively elsewhere in a nearly passive way [5]. Moreover, as opposed to conventional BSs, RISs can be fabricated to be extremely lightweight (hundreds of grams) and consume as little as tens of milliwatts, making them suitable for being rapidly deployed in beyond 5G/6G networks at affordable costs.
Figure 1 The RIS-based 3D communication environment with UAVs, HAPs, and FWAs.
While the opportunities of RIS-aided A2G networks are well understood, several design aspects still need to be addressed to unlock their full potential. Indeed, the joint optimization of the RIS and BS configuration, as well as the aerial device trajectory, require an adaptive level of control and signaling able to cope with the degree of mobility of the specific application scenario, including unpredictable movements typical of aerial objects, corresponding maneuvering, and a limited power budget available on board. All such aspects pose strict limitations on the A2G control architecture, making its design extremely challenging.
In this article, we identify the main optimization aspects to be considered when designing an RIS-aided A2G network by describing its main characteristics and corresponding state-of-the-art (SoA) solutions. We further shed light on technical challenges and open research directions and characterize two relevant operating scenarios. We finally deliver the message that while joint exploitation of RIS energy-efficient beamforming capabilities and flexible and agile aerial communications requires a careful design of the control architecture, it might significantly boost the overall communication performance.
RISs can be integrated in the context of aerial communications in two different ways, namely, 1) terrestrial RISs, i.e., by mounting them on the facades of buildings to assist in the communication to/from the flying object, and 2) aerial RISs, i.e., by employing them as substitutes to bulky active components such as conventional BSs on board the flying device [6]. However, while deploying terrestrial RISs can help alleviate the total power consumption, it may still limit the operation range of the flying object owing to the necessary active and heavy devices on board for communication purposes, whereas by allowing the UAV to carry an inherently lightweight and passive device on board such as an RIS, nearly all the available power can be devoted to enlarging the range of operation while simultaneously achieving highly selective beamforming. The RIS itself might even be (partially) powered by harvesting power from the incoming signals, which further demonstrates the feasibility of aerial RISs [7]. Moreover, terrestrial RISs can only achieve 180° reflection angles, compared with the full-angle 360° of aerial RISs, which allows the effective coverage of the intended 2D target area. Therefore, we focus on aerial RISs as depicted in Figure 1 and describe the main characteristics, involved opportunities, and corresponding challenges.
The link budget of an A2G network is directly proportional to the signal wavelength and the number of RIS elements and inversely proportional to the distance to the user [8]. In this regard, the use of millimeter-wave (mm-wave) or terahertz (THz) and the limited payload that UAVs can carry point to the need for optimizing the location of the aerial platform to obtain the best possible propagation conditions to the ground users in terms of increased LoS probability and limited path loss. Note that this problem can be mitigated by deploying RIS-equipped UAV swarms, i.e., a group of aerial RISs that cooperate together [9]. However, such cooperation comes at a significantly increased cost in control overhead and complexity.
On the other hand, beamforming at the RIS can increase the received power at the intended location up to a maximum theoretical factor equal to the square of the number of elements. To take full advantage of such tremendous capabilities, advanced 3D passive beamforming algorithms and techniques at the RIS are needed, especially in the multiuser scenario, which requires addressing complex nonconvex problems.
Moreover, current 5G BSs are usually equipped with a multiantenna array, which endows them with active beamforming capabilities. The optimization of both active and passive beamforming vectors is intrinsically coupled with the optimization of the position of the flying device and its evolution over time (i.e., its trajectory). Such joint optimization is in general intractable and requires one to resort to suboptimal approaches or heuristics. In particular, an efficient approach is decoupling the two problems in an alternating fashion. The trajectory optimization is efficiently solved via machine learning (ML) tools such as deep networks, while the passive beamforming optimization problem can be tackled via successive convex approximation or semidefinite relaxation (SDR) [10]. In the multiuser scenario, one can alternatively schedule users with a time division multiple access scheme [11] or, to improve fairness and robustness, consider max–min approaches thereby assigning to each user a given figure of merit, such as, e.g., the signal-to-noise ratio (SNR) or the bit error rate, and aiming to maximize the worst figure among the target users [12].
The core of the joint trajectory and beamforming optimization in A2G networks is in the acquisition and tracking of channel state information (CSI), relative to both the channel from the BS to the flying device and from the latter to the target user. However, the passive nature of RISs poses several constraints on the applicability of classical pilot-based channel estimation techniques [13]. In addition, in real-life scenarios, the users have a certain degree of mobility, which presents the highly challenging problem of tracking the evolution of the channel vectors over time. In this regard, it is essential to have accurate channel modeling that depends on a few key system parameters. While this may impact the system performance, it may also alleviate the task of obtaining CSI. Indeed, thanks to the high LoS probability, the channel vectors are mainly characterized by the distance traveled by the signal and the angle of arrival/angle of departure at the various entities in the network [14]. It is thus possible to realize efficient CSI acquisition in A2G networks by tracking the location of both the users and the flying device, which is feasible with a sufficient degree of control signaling. Moreover, note that while accurate channel modeling is essential to progress RIS-aided A2G networks and assess their performance, we focus on how CSI and other relevant system information can be used to design enhanced A2G control architectures.
Motivated by such considerations, in the following, we dig deeper into the design of an enhanced A2G control architecture to optimize the system performance while keeping the associated overhead to a minimum.
As depicted on the left-hand side of Figure 2, SoA control architectures for A2G networks in the case of UAVs are typically divided into two separate layers: ground control station and aerial platform control. The former is physically located in the terrestrial network, and it consists of a processing unit that, given a set of policies and quality of service (QoS) requirements, jointly optimizes the UAV trajectory and both the active and passive beamforming at the BS and at the RIS, respectively. Moreover, as described previously, it deals with acquiring CSI and extracting the associated relevant channel parameters, whereas the aerial platform control is given by the onboard UAV controller and the RIS controller. While the former is dedicated to the maneuvering of the UAV, the latter triggers RIS settings (i.e., predefined phase shifts). Note that such control architecture is generic and applicable regardless of the type of flying object involved. Moreover, the specific movement characteristics of the employed flying object are embedded into the maneuvering and sensor data, which are collected by the onboard controller and sent to the onboard intelligence module, as explained below.
Figure 2 The standard communication and control framework for the RIS-aided UAV scenario, with additional control enhancement modules and interfaces enabling advanced A2G control.
A widely adopted assumption in SoA control architectures is to consider the UAV in a predefined location in space with negligible orientation and position variations during the communication phase, while its position is updated only within the displacement phase. However, such an assumption does not hold in practical scenarios, wherein the UAV maneuvering and several atmospheric phenomena can change the position and the orientation of the UAV even during the communication operations, leading SoA solutions to be potentially inefficient—or even unfeasible—to operate in realistic conditions. Indeed, as the UAV is hovering at a certain altitude, its motion is influenced by a deterministic component due to the intentional maneuvering of the UAV, i.e., following a predefined trajectory, and a random component, due to unpredictable factors such as atmospheric conditions including wind, rain, and humidity, imprecise maneuvering, nonideal UAV instrumentation, etc. Such movements result in translations and rotations of the surface of the onboard RIS, leading to misalignment of the transmit and reflected beams. Therefore, the RIS-reflected signal may be directed toward unwanted locations in space and potentially generate destructive interference at nonintended users. Moreover, the highly directive nature of mm-wave beamforming at the RIS exacerbates this effect, potentially resulting in the loss of connectivity at the user side. Specifically, the received SNR may degrade by up to ∼25 dB [15].
To get the most from A2G networks in a practical scenario, the mitigation of UAV mobility effects on the QoS is a key point. It is thus essential to design enhanced control architectures enabling a transmission optimization tightly coupled with the UAV mobility pattern. Indeed, UAVs are equipped with different sensors, such as a gyroscope, compass, global positioning system, etc., that provide the UAV controller with motion feedback enabling hovering control and stabilization. However, due to the separation of the UAV and RIS controllers, information on UAV movements is only partially considered during the system optimization phase, i.e., only the nominal position of the UAV is considered to perform CSI acquisition and joint trajectory and beamforming optimization, while information such as real-time maneuvering instructions and UAV sensors’ output is typically neglected. Figure 2 shows the framework that we envision to enable interaction between the UAV and the control architecture. In particular, on the right-hand side, we consider two different logical modules forming the control enhancement, namely, onboard intelligence and in-network intelligence, which are located on the UAV and at the network side, respectively. These modules monitor and process information from the drone and the network and cooperate to enhance control. The onboard module interacts both with the UAV and the RIS controllers. It plays the fundamental role of collecting mobility information from the UAV and integrating it into the communication optimization process. Depending on the considered key performance indicators (KPIs), the latter may implement different types of algorithms to define the RIS configuration, ranging from artificial intelligence (AI) to conventional optimization tools [7], [12], [15]. Interestingly, the in-network intelligence module communicates both with the onboard module and with the standard communication optimization module. Our proposed novel framework is thus capable of blending together the various conventionally separated system entities and effectively utilizing all the available precious information to suitably optimize both RIS and BS parameters such as, e.g., beamforming configurations and transmit power at the BS.
To describe our envisioned enhanced control architecture, we first identify the key aspects that significantly affect its design. As depicted in Figure 3, an A2G network can explore three main directions: 1) the total available power budget; 2) the chosen reconfiguration rate, i.e., the rate at which the key system parameters—such as the UAV position, the beamforming vectors at the BS and the RIS, the CSI, etc.—need to be updated; and 3) the maneuvering control, i.e., identifying which entity in the system is controlling the UAV and how this interacts with the rest of the network. In the following, we describe the impact of each one of the aforementioned orthogonal directions, which are tightly coupled with the chosen hardware to be employed at both the UAV and the RIS, and the desired application scenario, with its associated physical propagation conditions and QoS requirements.
Figure 3 The design choices for the A2G enhanced control architecture.
Power availability: The specific UAV hardware directly determines the available power budget on board and the percentage used for flying or hovering. In addition, the UAV may be proprietary of the network operator or be outsourced from a third party to lower capital expenditures. In the first case, all the available power is dedicated to optimizing the underlying A2G network, whereas in the second case, the UAV movements are independently controlled by an external entity. Indeed, UAVs nowadays are used for several different applications, e.g., package delivery or surveillance. Thus, they can be equipped with RISs to be opportunistically used by the network. On the other hand, the size and weight of the RIS affect the total power consumption, thus limiting the range of the UAV, while its number of antenna elements defines the degree of computational power and overhead at the RIS controller and at the network side, e.g., to determine the passive beamforming configuration. In addition, the chosen application scenario plays a fundamental role in total power availability. In particular, the number of users and their mobility behavior, and the propagation conditions, such as the presence of obstacles or unwanted meteorological phenomena, are key aspects to be carefully considered. Moreover, the UAV needs to be recharged in appropriate charging stations, whose location constrains the total power availability in the environment. Such aspects directly impact the joint beamforming and trajectory optimization of the RIS and the UAV, respectively, and thus imply a given computational power. The latter is also directly influenced by the required QoS: in disaster situations, the focus is on sending emergency time-critical signals that demand low latency and low rate, thus giving rise to low computational power, whereas less critical applications such as data streaming may accept a larger latency but require a higher rate, which is typically associated with increased computational power. In this regard, we identify three categorizations, namely, unconstrained, feasible, and limited power availability.
Reconfiguration rate: The choice of reconfiguration rates enables different QoS guarantees and overall achievable communication rate. We determine three degrees of increasing reconfiguration rate dubbed as intermittent, regular, and frequent. For low reconfiguration rates, the system can devote most of the available time to sending/receiving data to/from the users. However, the available information from the UAV sensors, user feedback, and the CSI acquisition is only sporadically updated. As a result, the performance might be acceptable only for low-mobility scenarios or when the atmospheric conditions are ideal, whereas for a frequent reconfiguration rate, the available information is continuously updated, and as a result, the communication quality can be generally maximized even under high user mobility and strong meteorological perturbations. However, this implies an increased overhead that negatively affects the overall data rate. An intermediate solution is given by regular reconfiguration rates, which is the case of robust optimization algorithms, i.e., schemes that employ statistical channel and perturbation information rather than costly instantaneous CSI.
Maneuvering: The UAV maneuvering, including the evolution of its trajectory over time, is handled either by the network itself, i.e., it is coupled with the optimization of the system, or by an external operator as, e.g., in disaster situations, in which the information about the UAV position is feedback based from the control device. In such scenarios, the network can access useful information from the UAV sensors to accurately track the CSI as the UAV moves. In the second case, the UAV maneuvering might be under the control of an external entity and is thus disjoint from the rest of the system optimization. Hence, the CSI and the UAV position must be regularly estimated to keep an acceptable level of performance. Note that the focus of this article is not the design of such optimization algorithms or CSI acquisition schemes but rather to identify the need for an enhanced control architecture that suitably manages their interplay as to minimize overhead and optimize the communication performance.
Hereafter, we describe a practical implementation of our proposed enhanced control architecture considering two relevant case studies when RIS technology is in place.
Case study: static UAV: The relevant scenario of interest is depicted in Figure 4. Assume that due to low mobility of ground users, a UAV is in a fixed location in space, which is only seldom updated. However, due to adverse atmospheric conditions (e.g., wind, rain, and humidity), the UAV is subject to unwanted perturbations, which result in undesired roll, yaw, and pitch of the surface of the onboard RIS. Such problems are further exacerbated in A2G networks since the onboard RIS performs instantaneous passive signal reflection. UAV movement counteractions are automatically taken, but still orientation oscillations or location perturbation may result in an instantaneous RIS misconfiguration, leading to misalignment of the reflected beams and degraded overall achievable rate. Remarkably, it has been recently shown that it is possible to guarantee an acceptable level of performance in the target area of influence by suitably adjusting the beamforming configuration as a function of the second-order statistics of such perturbations [15]. Such an adaptation of the system configuration is enabled by our proposed enhanced control architecture. Indeed, the measurements of the instantaneous roll, yaw, and pitch are collected by the UAV controller and sent to the UAV data processing module that extracts or predicts (e.g., using AI) the relevant statistics. This information is then used upon request to update the current CSI and optimize the communication to/from the UAV. In particular, the beamforming configuration can be optimized on the basis of the current perturbation statistics via 1) conventional mathematical tools such as SDR, 2) training a ML model that learns how to adapt the beamforming configuration to the varying atmospheric conditions, or 3) designing an online AI learning algorithm. The choice of optimization method determines the computational power expenditure, which has a direct impact on the total power availability. As a result, our proposed enhanced control architecture can be designed to have a feasible or unconstrained power availability.
Figure 4 A practical scenario with a static UAV subject to undesired displacement and oscillation due to atmospheric phenomena. The output of the UAV sensors is collected and processed to statistically characterize the UAV mobility and obtain associated relevant CSI. The RIS configuration is optimized accordingly to generate reflected signal paths that are robust to the UAV displacements and preserve the communication quality. The control design (right-hand side) suits a scenario with doable or unconstrained power availability, frequent or regular reconfiguration rate, and coupled or feedback-based maneuvering.
As the UAV is in a fixed position, the RIS and BS precoders can be updated only when needed, i.e., when the perturbation statistics evolve due to a change of atmospheric conditions. Our proposed enhanced control architecture provides two ways to deal with such a scenario, which are characterized by an intermittent and regular reconfiguration rate, respectively. A first approach is to exploit feedback monitoring: users in the service area can periodically send updates to the network with the current perceived QoS. The in-network intelligence devoted to the CSI acquisition issues a statistics update request to the UAV data processing module if relevant changes in the QoS are detected.
A second approach is based on proactively sending updates on the UAV perturbation statistics when a relevant change is detected from the UAV data processing module to the CSI acquisition module. The latter then decides whether the relevant system parameters, such as the beamforming configurations, should be updated. This approach leads to higher power consumption at the UAV due to continuous monitoring and increased communication overhead. On the other hand, it allows quick reactions to varying environmental conditions and minimization of network downtime.
Last, note that both aforementioned approaches can be realized under coupled or feedback-based maneuvering. Indeed, in such cases, the UAV is controlled by the network and can thus retrieve useful information from the UAV controller, whereas a disjoint implementation would make it unfeasible to extract statistical information on the unwanted perturbations of the UAV position.
Case study: nomadic UAV: We consider the practical case where a UAV is subject to desired movements in space, i.e., following a predetermined trajectory, as depicted in Figure 5. For simplicity, we neglect user mobility to focus on providing coverage enhancement within a given target service area. The drone movements lead to a continuous position and orientation change of the onboard RIS. Such effect, if not properly addressed, could lead to severe beam misalignment and potentially to complete service disruption. Therefore, to maintain a stable connection, a continuous adaptation of both the RIS configuration and the BS precoder is required.
Figure 5 A practical scenario with a nomadic UAV subject to intentional maneuvering operations with corresponding displacement and orientation changes. The maneuvering operation and the output of the UAV sensors are processed and used to track and predict the UAV position and orientation in time. The RIS configuration is optimized according to the UAV motion to avoid connectivity disruption. The control design (right-hand side) suits a scenario with limited or doable power availability, frequent or regular reconfiguration rates, and coupled or feedback-based maneuvering.
Referring to the degrees of freedom highlighted in the section “Key Design Aspects,” we consider the UAV maneuvering to be either coupled, i.e., the UAV trajectory is imposed by decisions taken at the network side and therefore jointly optimized with the beamforming strategy, or feedback-based, i.e., the UAV is controlled by an external operator (e.g., a member of first responder teams), and therefore, its movements are not perfectly known to the network.
In the case of coupled maneuvering, the communication optimization module can compute in advance the RIS and the BS beamforming configuration according to the UAV trajectory evolution. Meanwhile, thanks to the UAV sensors data, the onboard data processing module can track the effective trajectory evolution, compare it against the desired one, and feedback information in case of divergence (e.g., nonidealities of the UAV controller, wind, etc.), thus enabling suitable adjustment of both the beamforming and the maneuvering and improving the overall reliability and robustness of the system. Depending on the power availability and variety of sensors equipping the UAV, the onboard data processing module can be further exploited. For example, visual information from cameras could be used to perform object detection and reveal potential obstruction (e.g., buildings, trees, or debris) and adapt the UAV trajectory accordingly.
In the case of feedback-based maneuvering, the UAV movements are known at the network side only a posteriori. Therefore, to prevent the communication optimization from lagging behind the UAV movements, a trajectory prediction strategy could be applied to the onboard data processing module. Such prediction could take advantage of both the sensor data and the maneuvering feedback, while the trajectory forecast could be sent to the network side to compute the optimal beamforming configuration in advance, so as to enable a more accurate transmission adaptation to the UAV movement.
The need for continuous communication adaptation, specific of the nomadic UAV scenario, results in a control architecture design that has to sustain a relatively high reconfiguration rate, i.e., regular or frequent. This implies a potentially high control overhead to frequently transmit the RIS configuration to the RIS controller. Nonetheless, smart overhead reduction strategies can be implemented thanks to the proposed enhanced control architecture. Indeed, the reconfiguration can be avoided as long as the QoS is within the desired level, despite the movement of the UAV. Therefore, a rate adaptation strategy could be considered to trigger the reconfiguration only when needed, e.g., based on the users’ feedback monitoring, thus minimizing the communication overhead. Moreover, the RIS configuration typically exhibits regular and periodic patterns over the surface elements. This peculiarity can be exploited by advanced encoding and decoding techniques, e.g., autoencoders, that could be easily implemented in network (encoding) and on board (decoding) to further reduce the overhead. Such an additional feature would stress the computational load of the onboard intelligence module and in turn reflect on the power availability requirements. As a result, this scenario is characterized by limited or feasible power availability.
The performances of our proposed enhanced control architecture in the application scenario described in the section “RIS-Aided A2G Case Studies for 6G” is hereafter summarized, for the case of feedback-based maneuvering, i.e., in the presence of an external UAV operator. We consider an A2G network including one single-antenna BS with 16 antenna elements, one UAV equipped with a 10 × 10 element squared RIS, and a target single-antenna user located at a distance of 70 m from the BS, whose height is set to 10 m. We assume the operator moves the UAV following a trajectory encircling the user with a radius of 25 m and an altitude of 30 m and with a given variable speed. The working frequency is set to 30 GHz, the transmit power at the BS is fixed to 24 dBm, the transmission time window is set to 50 ms, and the noise spectral density is assumed to be –80 dBm. Moreover, we adopt the channel model in [15].
In Figure 6, we compare two different schemes, namely, our proposed enhanced control architecture dubbed as adaptive and the standard SoA control framework, namely, fixed. The former implements an adaptive algorithm, thereby triggering an RIS reconfiguration whenever the received SNR falls below a given threshold, whereas the considered standard fixed control framework consists in updating the RIS configuration at periodic intervals, according to a predefined reconfiguration rate. We vary the UAV speed from 5 to 50 km/h and consider two different choices of reconfiguration rate, denoted as frequent in solid lines and regular in dashed lines, corresponding to reconfiguration rates of 1 and 2 Hz, respectively, in the case of standard fixed-rate control, and to an SNR threshold of 40 and 20 dB, respectively, in the case of the proposed adaptive control. The chosen reconfiguration rates for the two considered schemes are summarized in Table 1. We then evaluate the achievable rate of both schemes and the associated reconfiguration overhead, which is computed as the percentage of time spent exchanging control messages with respect to the total transmission time window. In the case of the standard fixed control framework, the overhead is obtained by simply multiplying the reconfiguration rate by the transmission time window, whereas for the proposed adaptive scheme, we consider the summation of all the triggered RIS reconfigurations in the transmission time window, which are caused by the SNR falling below the threshold.
Figure 6 A comparison of the achievable rate and the control overhead in a nomadic UAV scenario obtained with standard periodic RIS reconfiguration rate, dubbed as fixed, and smart adaptive reconfiguration enabled by the enhanced control architecture, dubbed as adaptive, versus the UAV speed. Frequent and regular stand for reconfiguration rates of 1 and 2 Hz, respectively, for the fixed reconfiguration scheme and for SNR thresholds of 30 and 20 dB, respectively, for the adaptive reconfiguration scheme.
Table 1 The reconfiguration rates for the proposed adaptive method and standard periodic control for frequent or regular reconfigurations.
Thanks to our proposed control architecture, the in-network intelligent modules receive feedback from the UAV maneuvering, the UAV sensors, and the user-perceived QoS and use it to adapt the system reconfiguration rate to the current UAV mobility rate and effective trajectory. In particular, as shown on the right-hand side, our proposed scheme adapts the CSI and RIS beamforming strategy more frequently for an increasing UAV speed. On one hand, this generates higher overhead, which is shown as a percentage of the total available transmission time. On the other hand, the data rate is kept high and quasi-constant for the case of frequent reconfiguration rate thanks to the up-to-date CSI, which in turn leads to high SNR.
In contrast to this, the standard architecture does not have access to the aforementioned UAV and QoS status information and thus uses a predetermined refresh rate to perform CSI acquisition and RIS beamforming optimization, as is typically the case for directional communications such as mm-wave. In this case, the communication overhead is constant versus the UAV speed, while the data rate monotonically decreases. However, at low UAV speeds, the standard scheme obtains higher data rates at the cost of an increased overhead as compared with the proposed enhanced scheme.
Last, we remark that in the above, we have showcased the potential benefits of implementing an enhanced A2G control architecture in a specific relevant use case and under a specific channel model, in the hope that this may foster future research to provide extensive and quantitative performance evaluations in several realistic scenarios.
Aerial communications are opening a new research direction to enable a 3D mobile networking paradigm expected to be effective in the 6G landscape. However, a number of daunting challenges need to be addressed before the dream of cost-effective flying mobile stations solutions can be reached, in particular, with respect to the design of the control architecture.
In this article, we analyzed the novel concept of RIS-aided aerial communications and shed light on some of its potential use cases, optimization aspects, and challenges. Specifically, we focused on aerial RISs and argued that a carefully designed enhanced control architecture is essential to take full advantage of the RIS 3D passive beamforming capabilities and the flexibility of flying devices. In contrast to existing SoA frameworks, our envisioned enhanced control architecture is able to bridge together several conventionally isolated entities and exploit useful information generated by both the flying device sensors and the user feedback on the perceived QoS. This allows for obtaining high-performance networks that adapt the system configuration to varying environmental conditions while minimizing the associated control overhead.
We analyzed two practical scenarios of interest wherein the underlying A2G network benefits from our proposed architecture and quantified its performance in a relevant case study. Our results show that 1) our proposed adaptive scheme is able to adapt the CSI and RIS beamforming strategy as required according to varying UAV speeds, 2) the adaptability comes at a nonnegligible reconfiguration overhead cost of about 5% to 15%, and 3) the proposed adaptive solution successfully manages to keep the data rate at low degradation percentages (<10%) for the UAV speed range considered (5 to 50 km/h).
This work was supported by the European Union H2020 Project RISE-6G under Grant 101017011.
Francesco Devoti (francesco.devoti@neclab.eu) received his B.S. and M.S. degrees in telecommunication engineering and his Ph.D. degree in information technology from Politecnico di Milano in 2013, 2016, and 2020, respectively. He is currently a senior research scientist in the 6G Network Group at NEC Laboratories Europe, 69115 Heidelberg, Germany. His research interests include reflective intelligent surfaces, mm-wave technologies in 5G and 6G networks, and network slicing. He is a Member of IEEE.
Placido Mursia (placido.mursia@neclab.eu) received his B.Sc. and M.Sc. (with honors) degrees in telecommunication engineering from Politecnico di Torino in 2015 and 2018, respectively. He obtained his Ph.D. from Sorbonne Université of Paris, at the Communication Systems Department of EURECOM in 2021. He is currently a research scientist in the 5GN group at NEC Laboratories Europe, 69115 Heidelberg, Germany. His research interests are convex optimization, signal processing, and wireless communication. He is a Member of IEEE.
Vincenzo Sciancalepore (vincenzo.sciancalepore@neclab.eu) received his M.Sc. degree in telecommunications engineering and telematics engineering in 2011 and 2012, respectively, and in 2015, he received a double Ph.D. degree. Currently, he is a principal researcher at NEC Laboratories Europe, 69115 Heidelberg, Germany, focusing his activity on RISs. He is an editor of IEEE Transactions on Wireless Communications. He is a Senior Member of IEEE.
Xavier Costa-Pérez (xavier.costa@ieee.org) is a research professor at ICREA, scientific director at the i2Cat R&D Center, and head of 6G networks R&D at NEC Laboratories Europe, 08034 Barcelona, Spain. He has served on the program committees of several conferences, has published at top research venues, and holds several patents. He received his M.Sc. and Ph.D. degrees in telecommunications from the Polytechnic University of Catalonia in Barcelona. He is a Senior Member of IEEE.
[1] M. Mozaffari, W. Saad, M. Bennis, Y. H. Nam, and M. Debbah, “A tutorial on UAVs for wireless networks: Applications, challenges, and open problems,” IEEE Commun. Surveys Tuts., vol. 21, no. 3, pp. 2334–2360, 3rd Quart. 2019, doi: 10.1109/COMST.2019.2902862.
[2] A. Albanese, V. Sciancalepore, and X. Costa-Pérez, “First responders got wings: UAVs to the rescue of localization operations in beyond 5G systems,” IEEE Commun. Mag., vol. 59, no. 11, pp. 28–34, Nov. 2021, doi: 10.1109/MCOM.101.2100273.
[3] E. Calvanese Strinati and S. Barbarossa, “6G networks: Beyond Shannon towards semantic and goal-oriented communications,” Comput. Netw., vol. 190, May 2021, Art. no. 107930, doi: 10.1016/j.comnet.2021.107930.
[4] S. Alfattani et al., “Aerial platforms with reconfigurable smart surfaces for 5G and beyond,” IEEE Commun. Mag., vol. 59, no. 1, pp. 96–102, Jan. 2021, doi: 10.1109/MCOM.001.2000350.
[5] E. C. Strinati et al., “Reconfigurable, intelligent, and sustainable wireless environments for 6G smart connectivity,” IEEE Commun. Mag., vol. 59, no. 10, pp. 99–105, Oct. 2021, doi: 10.1109/MCOM.001.2100070.
[6] X. Pang, M. Sheng, N. Zhao, J. Tang, D. Niyato, and K.-K. Wong, “When UAV meets IRS: Expanding air-ground networks via passive reflection,” IEEE Wireless Commun., vol. 28, no. 5, pp. 164–170, Oct. 2021, doi: 10.1109/MWC.010.2000528.
[7] Q. Zhang, W. Saad, and M. Bennis, “Reflections in the sky: Millimeter wave communication with UAV-carried intelligent reflectors,” in Proc. IEEE Global Commun. Conf. (GLOBECOM), 2019, pp. 1–6, doi: 10.1109/GLOBECOM38437.2019.9013626.
[8] S. Alfattani, W. Jaafar, Y. Hmamouche, H. Yanikomeroglu, and A. Yongaçoglu, “Link budget analysis for reconfigurable smart surfaces in aerial platforms,” IEEE Open J. Commun. Soc., vol. 2, pp. 1980–1995, Aug. 2021, doi: 10.1109/OJCOMS.2021.3105933.
[9] B. Shang, R. Shafin, and L. Liu, “UAV swarm-enabled aerial reconfigurable intelligent surface (SARIS),” IEEE Wireless Commun., vol. 28, no. 5, pp. 156–163, Oct. 2021, doi: 10.1109/MWC.010.2000526.
[10] A. Khalili, E. M. Monfared, S. Zargari, M. R. Javan, N. Mokari, and E. A. Jorswieck, “Resource management for transmit power minimization in UAV-assisted RIS HetNets supported by dual connectivity,” IEEE Trans. Wireless Commun., vol. 21, no. 3, pp. 1806–1822, Mar. 2021, doi: 10.1109/TWC.2021.3107306.
[11] Y. Li, C. Yin, T. Do-Duy, A. Masaracchia, and T. Q. Duong, “Aerial reconfigurable intelligent surface-enabled URLLC UAV systems,” IEEE Access, vol. 9, pp. 140,248–140,257, Oct. 2021, doi: 10.1109/ACCESS.2021.3119268.
[12] H. Lu, Y. Zeng, S. Jin, and R. Zhang, “Aerial intelligent reflecting surface: Joint placement and passive beamforming design with 3D beam flattening,” IEEE Trans. Wireless Commun., vol. 20, no. 7, pp. 4128–4143, Jul. 2021, doi: 10.1109/TWC.2021.3056154.
[13] C. Hu, L. Dai, S. Han, and X. Wang, “Two-timescale channel estimation for reconfigurable intelligent surface aided wireless communications,” IEEE Trans. Commun., vol. 69, no. 11, pp. 7736–7747, Nov. 2021, doi: 10.1109/TCOMM.2021.3072729.
[14] H. Jiang et al., “A novel 3D UAV channel model for A2G communication environments using AoD and AoA estimation algorithms,” IEEE Trans. Commun., vol. 68, no. 11, pp. 7232–7246, Nov. 2020, doi: 10.1109/TCOMM.2020.3011716.
[15] P. Mursia, F. Devoti, V. Sciancalepore, and X. Costa-Pérez, “RISe of flight: RIS-empowered UAV communications for robust and reliable air-to-ground networks,” IEEE Open J. Commun. Soc., vol. 2, pp. 1616–1629, Jun. 2021, doi: 10.1109/OJCOMS.2021.3092604.
Digital Object Identifier 10.1109/MVT.2023.3274329