EP4716994A1 - Coordinated reconfigurable intelligent surfaces - Google Patents

Abstract

Techniques and methods are described to coordinate multiple reconfigurable intelligent surfaces (RISs), so as to determine a communication path between a transmitter and a receiver communicating via a channel. In order to determine the communication path, channel information that relates the channel between the transmitter and the receiver is obtained. Further, for each RIS of the multiple RISs, a set of RIS parameters is obtained, where the set includes multiple parameters related to said RIS. The sets of RIS parameters and the channel information are then used to determine the communication path that includes two or more RISs.

Description

Coordinated Reconfigurable Intelligent Surfaces The present disclosure relates generally to wireless communication, and in particular to coordination of reconfigurable intelligent surfaces. BACKGROUND The increasing demand on data rates from the massive number of devices motivates the need to develop novel system architectures that are both energy and spectrally efficient. For the past years, state of the art research has focused on leveraging large-scale Multiple-Input-Multiple- Output (MIMO) systems, such as massive and millimeter wave (mmWave) MIMO at the base stations (BSs) and mobile users. To further improve the coverage and the energy efficiency of these systems, and hence improve the performance of wireless data transmission systems, reconfigurable intelligent surfaces (RISs) have been recently proposed and attracted substantial interest. A single RIS has a large numbers of small reflecting units which are jointly adjusted to reconfigure the environment of the wireless signal transmission (wireless environment). In most works, a single RIS is used in a system setup, where the reflection coefficients of the reflecting units of one RIS are adapted to reflect a wireless signal incident on one RIS in a desired direction.. Still, there are promising setups to further exploit the advantages of the RIS by using multiple RISs in a system. The RISs in one RIS system as well as RISs belonging to different RIS systems have an effect on each other, asking for coordinating the RISs. However, there is little work to exploit such coordination of multiple RIS. One example of a use case for coordination is path selection, where several parameters and overall system performance should be considered. In order to exploit the overall system performance and to account for multiple parameters characterizing the wireless environment in the present of multiple RISs, there should be coordination between the RISs, which remains a challenging task. SUMMARY Methods and techniques are described herein for facilitating a reliable and robust communication and sensing data. For that purpose, the present disclosure provides methods and techniques to coordinate multiple RIS mediating wireless communication between a base station (BS) and one or more terminals, so as to improve the performance of wireless transmission systems. The invention is defined by the independent claims. Some exemplary implementations are provided by the dependent claims. For example, a control device is provided for coordination of a plurality of reconfigurable intelligent surfaces, RISs, the control device comprising: a communication module configured to: obtain channel information related to a channel between a transmitter and a receiver; and obtain, for each RIS of the plurality of RISs, a set of RIS parameters including a plurality of parameters related to said RIS, and processing circuitry configured to: determine a communication path between the transmitter and the receiver based on the channel information and based on the sets of RIS parameters of the plurality of RISs, wherein the communication path includes two or more RISs of the plurality of RISs. The above mentioned circuitry may be any circuitry such as processing circuitry including one or more processors and/or other circuitry elements. Furthermore, a method is provided for coordination of a plurality of reconfigurable intelligent surfaces, RISs, the method comprising: obtaining channel information related to a channel between a transmitter and a receiver; obtaining, for each RIS of the plurality of RISs, a set of RIS parameters including a plurality of parameters related to said RIS, and determining a communication path between the transmitter and the receiver based on the channel information and based on the sets of RIS parameters of the plurality of RISs, wherein the communication path includes two or more RISs of the plurality of RISs. These and other features and characteristics of the presently disclosed subject matter, as well as the methods of operation and functions of the related elements of structures and the combination of parts and economies of manufacture, will become more apparent upon consideration of the following description and the appended claims with reference to the accompanying drawings, all of which form a part of this specification. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the disclosed subject matter. As used in the specification and the claims, the singular form of “a”, “an”, and “the” include plural referents unless the context clearly dictates otherwise. BRIEF DESCRIPTION OF DRAWINGS An understanding of the nature and advantages of various embodiments may be realized by reference to the following figures. FIG.1 is an illustration for RIS selection in two exemplary environments according to prior art: indoor (Fig.1A) and outdoor (Fig.1B). FIG.2 is an illustration of coordinating RISs (CRIS) according to the present disclosure. FIG.3 is a block diagram illustrating an exemplary control device performing coordinated RIS for determining a communication path. FIG.4 is a flow diagram illustrating exemplary steps performed by a control device performing steps of coordinating RIS to determine a communication path. FIG.5 is an illustration of a classical reinforcement learning (RL) with an agent observing the environment and performing action(s). FIG.6 is an illustration of a basic system model of the RIS network. FIG.7 is an illustration of the RL-based approach for coordinating RISs according to the present disclosure. FIG.8 is an illustration of using a base station BS to perform coordination of the RISs. FIG.9 is an illustration of coordination RISs belonging to different RIS networks. Like reference numbers and symbols in the various figures indicate like elements, in accordance with certain example implementations. DETAILED DESCRIPTION For purposes of the description hereinafter, the terms “end,” “upper,” “lower,” “right,” “left,” “vertical,” “horizontal,” “top,” “bottom,” “lateral,” “longitudinal,” and derivatives thereof shall relate to the disclosed subject matter as it is oriented in the drawing figures. However, it is to be understood that the disclosed subject matter may assume various alternative variations and step sequences, except where expressly specified to the contrary. It is also to be understood that the specific devices and processes illustrated in the attached drawings, and described in the following specification, are simply exemplary embodiments or aspects of the disclosed subject matter. Hence, specific dimensions and other physical characteristics related to the embodiments or aspects disclosed herein are not to be considered as limiting unless otherwise indicated. No aspect, component, element, structure, act, step, function, instruction, and/or the like used herein should be construed as critical or essential unless explicitly described as such. Also, as used herein, the articles “a” and “an” are intended to include one or more items and may be used interchangeably with “one or more” and “at least one”. Furthermore, as used herein, the term “set” is intended to include one or more items (e.g., related items, unrelated items, a combination of related and unrelated items, and/or the like) and may be used interchangeably with “one or more” or “at least one”. Where only one item is intended, the term “one” or similar language is used. Also, as used herein, the terms “has”, “have”, “having”, or the like are intended to be open- ended terms. Further, the phrase “based on” is intended to mean “based at least partially on” unless explicitly stated otherwise. Reconfigurable intelligent surfaces (RISs) have recently been prospected as one of the key technologies to achieve smart radio environment for the sixth generation (6G) of wireless communication systems. Specifically, in order to create a true smart radio environment, one of the considered key approaches is to make the radio medium controllable. To this end, RISs, also known as intelligent reflecting surfaces (IRSs) or software-defined metasurfaces, have been developed. In particular, a RIS consists of nearly passive reflecting elements, which are programmable and controllable via a RIS controller, allowing the RIS to reflect and steer impinging signals toward desired directions. In order to realize this goal, the phase-shifts of the reflecting elements are adjusted, such that the directionality of the beam of scattered signals can be controlled. With this phase-tuning capability, coherent signal combining of reflecting signals from different elements of different RISs can be properly implemented, such that the multi-path received signals constructively add at the receiver side. So far, work on RIS-based smart radio environment have been focused on single RIS. However, multiple RIS can be considered to allow further reflecting and steering of impinging signals toward desired directions that may not be achievable by a single RIS. Currently, there is no work for the coordination of multiple-RIS systems. On the other hand, coordination is essential for several use cases. Path selection is one of these scenarios. In the following, the available technologies on path selection of multiple RIS systems is described, followed by a detailed discussion of coordinating multiple RIS of the present disclosure. However, note that coordination of multiple RISs is not limited to path selection. It can also be used for resource allocation, security, etc. Fig. 1 illustrates a RIS-selection-based transmission over multiple RISs, with an indoor propagation scenario being shown in Fig.1A, and an outdoor propagation scenario being shown in Fig.1B, respectively. Each RIS is equipped with N reflecting elements, with N being equal to 2 or larger. In the example shown in Fig.1A, N = 16, while N = 32 in Fig.1B. As shown in Fig.1A, a number of ^ RISs are placed in an indoor propagation environment, where each RIS has ^ reflecting elements. Here, ^_^^^and ^_^^^, respectively represent the distances between a source ^ to the ^th RIS, and between the kth RIS to a destination D, for ^ = 1,2, … , ^. Furthermore, small-scale fading coefficient of the ^-to-^th RIS channel and ^th RIS-to-D channel is denoted by respectively for ^ = 1,2, … , ^ and ^ = 1,2, … , ^. The index i refers to the i-th element of one RIS. The received signal-to-noise ratio (SNR) of the signal transmitted through only the ^th RIS (^_^) is calculated as: where ^^^ ^ is the phase difference for the ^th path. Considering the maximized SNR values ^max,1, ^max,2, … , ^max,k, an RIS selection is conducted over ^_^ RISs to choose the RIS, which has highest SNR, for transmission as ^^ ^_^ = max ^^max,1, ^max,2, … , ^max,k^ where ^^ ^_^ is the maximized received SNR for the RIS selection-based system in indoor environment, and ^_^^^,^ is the maximized SNR of the ^th path with proper phase adjustment. The generalized system model of the RIS selection-based outdoor communication system is illustrated in Fig.1B. In this model, ^_^ and ^_^ RISs are assumed, respectively, within the proximity of the transmitter and receiver. Here, the distances of ^-to-RIS ^, RIS ^ -to-RIS ^ (i.e. the distance between the k-th and l-th RIS), and RIS ^-to-D are respectively denoted by ^_^^^, ^_^^^^^ and ^^{^} _^^^ for ^ = 1,2, … , ^_^ and ^ = 1,2, … , ^_^. In this setup, ^_^ × ^_^ possible transmission paths are available for communication. By selecting the path that has the highest SNR, the overall cost can be reduced using a single RIS for each terminal. The maximized SNR is obtained for each path as ^ ^,^ ^_^^ . Therefore, RISs selection for outdoor environment is conducted over ^_^ × ^_^ paths as: where ^ ^ ^_^^ is the maximized received SNR for RIS selection-based systems in outdoor environment. Coordination in wireless communication systems have been investigated in the context of Multi- Access Point Coordination and Coordinated Multi-Point transmission and reception, but not for RIS systems. However, if coordination between RISs taken into account, the path selection and resource allocation can be made more efficient, because coordinated RISs represent more accurately the smart wireless environment. Otherwise, each RIS will select or act in a manner which is best for itself in isolation to its surrounding other RISs. However, in the presence of multiple RISs participating in the wireless communication, the overall system optimization should be considered. In the examples above, the SNR-based method is explained for the case of path selection. However, several other parameters should be considered for the optimum path, because only SNR may not be sufficient for selecting the optimum path. For example, the SNR can be high, but latency may also be high, or security may cause problems. Therefore, it may be advantageous to consider several parameters simultaneously for optimum path selection. In addition or alternatively, the usage of multiple RISs may be optimized not only for a single transmission, but rather the overall system performance may be considered in some embodiments. For example, if the RIS load was critical, and the selected RIS was used for other transmissions, interference may be created (generated) to the systems. Therefore, it may be advantageous to coordinate the RIS network with other RIS networks. The coordination between the RIS systems brings several advantages, and it can be used for several use cases, such as path selection, resource allocation, or the like. In the present disclosure, path selection may be enhanced by accounting for coordination between RISs. In multiple RIS systems, it may be desirable to select RISs (i.e. a set of RIS having two or more RISs among a plurality of RISs) representing an optimum path, while accounting for the overall smart wireless environment as well as for various parameters of the RISs. The present disclosure provides for such efficient selection mechanism, where multiple RIS are coordinated so as to select a suitable (e.g. an optimal) path. This scheme is referred to coordinated RIS (CRIS). The basic idea of RIS coordination is illustrated in Fig.2. Note that in Fig.2 thick black lines represent the optimum path between a source transmitter (Tx) and a target receiver (Rx), while black thin lines (short-dashed, long-dashed, dash-dotted, dash-double- dotted, dotted) represent coordination between the RISs. It is noted that source transmitter and target receiver are functional terms. Irrespectively of the device, a source transmitter is a transmitter that transmits a signal to a system of RIRs (multiple RISs), and target receiver is a receiver that receives the signal from the system of RISs. The source transmitter and the target receiver may themselves be a RIS (e.g. from another RISs system or the like). Fig.2 shows a plurality of reconfigurable intelligent surfaces (RISs), which are to be coordinated such that two or more RISs of the plurality of RISs form a communication path between the transmitter Tx and receiver Rx. In the example shown in Fig.2, the plurality of RISs include ten RISs 210 to 219. The number of RISs is not limited to ten, but may include any integer equal to or larger than two. In order to coordinate the RISs, channel information is obtained, with said channel information related to a channel between the transmitter and the receiver. In the example, said channel information is obtained for RISs 210 to 219. The channel information represents the wireless environment of the RISs. For example, the channel information includes location information of the transmitter and/or the receiver, respectively. The location information may be global position system (GPS) coordinates. For each RIS of the multiple RISs, a set of RIS parameters is obtained that include a plurality of parameters related to said RIS. For the coordination, the plurality of RIS parameters include one or more of : traffic load, security, channel quality, latency, reliability, energy efficiency, throughput, time of arrival (ToA), time of flight (ToF), time of transmission (ToT), time difference of arrival (TDoA), and received signal strength indicator (RSSI). RSSI refers to a value of a signal strength measurement of a reference signal. The reference signal may be a pilot signal, for example. In an exemplary implementation, the RSSI is included in the channel information obtained from the receiver Rx. The set of RIS parameters may include one or more of the above-listed parameters. The advantages of using these parameters are given below. ^ Traffic load: The RIS traffic load means the amount of data moving across a RIS at a given point of time. The result of increasing network traffic is latency, or the delay from input into a system to its outcome. Therefore, these parameters should be considered when the RISs are coordinated with each other. ^ Security: Some RIS can be more secure due to their locations. The respective location of the RIS may be given in terms of three-dimensional (3D) coordinates e.g. x, y, and z, or in terms of two-dimensional (2D) coordinates in which case the RIS location refers to a 2D plane. Also, there are several works which use RIS from both attack and defense perspectives. Based on the security of the RISs, a RIS can be selected/used. Some critical applications, such as military services, healthcare, or autonomous vehicles applications can use more secure RIS. ^ Channel quality: Channel quality between the RISs is a critical factor for path selection. This is because channel quality is the factor that directly affects the bit error rate (BER). ^ Latency: The parameter of the latency should be considered for the path selection. Lowering the latency is better for the performance. ^ Reliability: The parameter of the reliability should be considered for the path selection. The higher the reliability, the better the performance. ^ Energy efficiency: Energy efficiency is usually defined as the number of bits that can be sent over a unit of power consumption which is usually quantified by bits per Joule. The determining factor of energy efficiency for mobile devices is the power needed to transmit data. Energy-efficient wireless communication network design is an important and challenging problem and RIS coordination should consider energy efficiency. ^ Throughput: Throughput is the actual amount of data that is successfully sent/received over the communication link (i.e. downlink (DL) and/or uplink (UL)). Throughput is presented as kilobits per second (kbps), Megabits per second (Mbps), or Gigabits per second (Gbps), and can differ from bandwidth due to a range of technical issues, including latency, packet loss, jitter and more. These parameters (latency, packet loss, jitter) can be also considered individually. ^ ToA, ToF, and TDoA: time of arrival (ToA) is the absolute time instant when a radio signal emanating from a transmitter reaches a remote receiver. The time span elapsed since the time of transmission (ToT) is the time of flight (ToF). Time difference of arrival (TDoA) is the difference between ToAs. These parameters are important since they have a direct effect on the delay and latency of the system. ^ RSSI: The received signal strength indicator (RSSI) is an estimated measurement of how well a device can hear, detect, and receive signals from any wireless access point (AP) or Wi-Fi router. A value of the RSSI closer to 0 is stronger, and closer to –100 is weaker. For best performance, the RSSI should be as high as possible. Coordination can be done using only one parameter. However, using more than one parameter may be advantageous. This is because parameters may have relationships with each other and they can affect each other (parameter interaction). For example, when the latency is lowered, the reliability can decrease. Therefore, jointly learning the coordination/path from these two parameters would be advantageous. Also, the applications employing RISs such as military services and communication, healthcare, autonomous vehicles or the like require several properties to be considered jointly. For example, one RIS can have a line-of-sight (LOS) path. However, if the RIS traffic load is too high, it may not be useful and another RIS could possibly achieve better results. Therefore, considering several parameters may advantageously be considered to achieve an optimal communication path employing multiple RISs. The coordination can be performed using one or more of the above parameters (RIS parameters) included in the set or RIS parameters. The best path between the RISs can be found and the optimization can be made, based on one or more of these RIS parameters. In other words, a communication path between the transmitter Tx and the receiver Rx is determined based on the channel information and based on the sets of RIS parameters of the plurality of RISs. The communication path includes two or more RISs of the plurality of RISs. In the Example of Fig.2, the determined communication path between the transmitter and the receiver is formed by three RISs 210, 216, and 213, which belong to the plurality of RISs 210 to 219. In an exemplary implementation, the receiver (target) and / or the transmitter (source) is a wireless station and / or a RIS. For example, the target receiver may be a user equipment (UE) or station (STA), such as a mobile phone, a smartphone, a smart tablet, a laptop, or a personal computer (PC) or the like. Similarly, the source transmitter may be a user equipment (UE) such as a mobile phone, a smartphone, a smart tablet, a laptop, or a personal computer (PC) or the like. The source transmitter may also be a base station. In case of a PC, the PC may be wired or wire-less connected to a wireless router enabling a wireless communication connection to the RISs of the RIS system. It is noted that the above are mere examples, and any other device may be used as long as it may be configured for wireless communication. The RIS serving as the source transmitter TX may be RIS 210 in Fig.2. Likewise, the target receiver may be RIS 213 in Fig.2. Fig. 3 shows control device 300 that performs the coordination of the RISs, and comprises communication module 310 and processing circuitry 320. The communication module obtains the channel information that relates to the channel between the source transmitter Tx and the target receiver Rx. Further, the communication modules obtains, for each RIS of the plurality of RISs, a set of RS parameters that include a plurality of parameters related to said RIS. For example, in case of RIS parameter being RSSI, the RSSI may be obtained from pilot signals that the control device (e.g. one RIS of the multiple RISs) received from the source transmitter and the target receiver that transmitted said pilot signals (an example of a reference signal) to the control device. The control device then calculates RSSI based on the received pilots. Based on the channel information and the sets of RIS parameters, the processing circuitry determines a communication path between the transmitter Tx and the receiver Rx, with the communication path including two or more RISs of the plurality of RISs. The control device may be an access point (AP) or a RIS among the plurality of RIS. In other words, the coordination of the RISs is executed by an access point or one of the RISs of said plurality of RISs. In an exemplary implementation, the processing circuitry controls the communication module to provide path information regarding the communication path to the two or more RISs that are included in the communication path. In the example of Fig.2, the two or more RIS may be any of two or three RISs 210, 216, and 213. For example, the path information provided to a specific RIS of the two or more RISs includes identification of a RIS that precedes said specific RIS in the communication path and/or identification information of a RIS following said specific RIS in the communication path. In Fig. 2, for example, with RISs 210, 216, and 213 forming the communication path between the transmitter and the receiver, the specific RIS may be RIS 216 to which identification of RIS 210 is provided, with RIS 210 preceding the specific RIS 216. Likewise, in Fig. 2, the RIS following the specific RIS 216 is RIS 213 whose identification is provided to specific RIS 216. It is noted that the terms “preceding” and “following” are to be understood, for example, with reference to a direction of the communication path. Thereby, the term “direction” may refer to a direction in terms of a signaling direction, where one of the transmitter and the receiver may be the source or the destination, respectively. For example, in downlink communication, the transmitter is commonly considered as the source and the receiver is the destination. In this case, the direction may be defined from the transmitter to the receiver. As a result, RIS 210 precedes and RIS 213 follows the specific RIS 216 in Fig.2. In turn, in uplink communication, the direction would be reversed, and the communication path direction would be from the receiver to the transmitter. In this case, RIS 214 precedes and RIS 210 follows the specific RIS 216 in Fig.2. As discussed above, the communication path includes two or more RISs, enabling communication between the (source) transmitter and the (target) receiver, where some RISs may be closer or farther distant to the transmitter and/or the receiver than other RISs. For example, with the terms “closer” and “farther distant” having their common meaning in that they refer to a physical distance RIS 210 in Fig.2 is closer to the source transmitter Tx than RIS 216. Hence, RIS 216 is farther distant to the source transmitter Tx than RIS 210. Similarly, RIS 213 in Fig.2 is closer to the target receiver Rx than RIS 216. Further, RIS 210 is closer to the source transmitter than to the target receiver. Hence, the terms “closer” and” farther distant” refer to physical distances between different RISs and the source transmitter and/or the target receiver. This implies that the RIS system with its multiple RISs has position information, as well as one or more of the RISs of said RIS system has position information (e.g. GPS position). The position of a single RIS may then be used to determine the physical distance between said RIS and the source transmitter TX and/or the target receiver Rx. For example, with the position of the transmitter and/or the receiver known from GPS position information, the physical distance between a RIS and the TX and/or Rx may be calculated from the difference between the transmitter GPS position and the RIS GPS position, and respectively from the difference between the receiver GPS position and the RIS GPS position. As exemplified above, the terms “closer” and “farther distant” are understood in a comparative manner. In addition, identification of a RIS of the two or more RISs that is/are closest to the source transmitter is provided to the transmitter, with the identification pertaining to the communication path. Similarly, identification of a RIS of the two or more RISs that is closest to the receiver is provided to the receiver, with the identification pertaining to the communication path. It is noted that, even though even though said RIS beings to the two or more RISs of the determined communication path, said communication path may not yet be optimal. Rather, during optimizing the communication path (e.g. employing an RL-based machine learning (ML) as detailed further below), the identification of said closest RIS may be used while still determining the optimal communication path. In this respect, Fig.2 showing an optimum communication path that includes three RISs 210, 216, and 213, respectively (thick solid line) may not be understood as a limitation. Rather, it illustrates how among multiple possible paths between the source transmitter and the target receiver, mediated by the plurality of RISs 210 to 219, two or more RISs are determined / selected to form an optimal communication path. In the example of Fig. 2, the result of such optimization are the three RISs 210, 213, and 216. For mere illustration purposes, the communication path with RISs 210, 213, and 216 may be alternatively interpreted as being an intermediate communication path while still in the process of finding the optimal path. In other words, the thick line in Fig.2 may be alternatively viewed as snapshot of the communication path that may not have approached its best optimum yet. Rather, since the optimization is in general an iterative process, the determined communication path may be interpreted as optimal path at a particular iteration step. The provision of said identification may be performed by communication module 310 of control device 300 in Fig.3. For example, the RISs of a RIS system may each have a unique ID for that system. Such ID may be, for example, a positive integer equal to or larger than 0. Alternatively or in addition, a RIS ID may be a position of the RIS (e.g. a GPS coordinate), which may be suitable to distinguish the RISs as long as the GPS RIS positions are different. The communication module may provide the RIS ID to the transmitter and/or receiver by transmitting, for example, the path information that may include also RIS identification (e.g. RIS ID and/or RIS position) of the respective closest RIS to transmitter RX and/or receiver Rx. Referring to the example of Fig. 2 and understanding the arrangement of transmitter-RISs- receiver as a two-dimensional (2D) spatial arrangement for the purpose of illustration, the RIS that is closest to the transmitter is RIS 210 compared to RISs 216 and 213, while RIS 213 is closest to the receiver compared to RISs 210 and 216. Hence, in this example, the identification of RIS 210 would be provided to the transmitter, while the identification of RIS 213 would be provided to the receiver. The identification pertaining to the communication path means that the identification of the respective RIS is included (after selection) in the path information. It is noted that Fig.2 is merely illustrating a specific 2D spatial arrangement of the transmitter, the receiver, and the plurality of RISs with its two or more RISs determining the communication path. It is understood that the RISs are typically arranged in a three-dimensional (3D) space, with RISs having a 3D position (e.g.3D GPS position), respectively. The optimum path using the above defined parameters may be determined by means of machine learning (ML) or deep learning algorithms. Using ML can be advantageous, since there can be several parameters for coordination. An example ML strategy is further discussed below, where a trained machine-learning model is applied to determine a communication path. The ML-based strategy relies on a trainable network (e.g. neural network and its various kind of sub-classes (e.g. convolutional neural network etc.)), which is initially trained for providing a specific output upon providing input data as training data. Here, the input training data include channel information (e.g. channel vector) and outputting an interaction vector. These details are discussed below. Alternatively, a statistical algorithm may be used to determine a communication path. However, a best path may be found using signals in RISs. According to an exemplary implementation, the RIS coordination may be based on ML in conjunction with a reinforcement learning (RL) strategy. This RL strategy may be advantageous for the problem of RIS coordination, because it relies on interacting of the RISs with the environment, and may not require predefined data for determining the communication path. Also, the model is trained and tested simultaneously in the RL algorithm. A typical RL algorithm suitable for RIS coordination is illustrated in Fig.5. The RL algorithm assumes an environment and an agent that interact. As illustrated in Fig.5, an agent is in a state (e.g. initial or current state) and performs an action, which in turn affects the environment in that it takes a new state. With the change of the state upon the action, there is a reward associated, with the reward being a measure for how likely (i.e. probable) it is to change from a current state to new state, when the is taken. This reward is provided as input to the agent together with the state, based on which the agent takes another action. Hence, there is an exchange of information and hence feedback between the agent and the environment (interaction). Applying the agent-environment framework of Fig. 5 to the RIS coordination using the RL algorithm, the steps for how the RL is used for coordinating multiple-RISs for path selection may be summarized as follows: ^ The location information of the transmitter and receiver is indicated. The location information of the transmitter and receiver will be the input state (i.e. the state in Fig.5). This can be the distance of the transmitter and receiver to a single RIS, coordination information, etc.. Coordination information may include the GPS coordinates (e.g. of the transmitter and/or receiver and/or any of the RISs). Also, a multipath signature between the transmitter and receiver can be used as an input. Note that different inputs, such as mobility information of the RIS, the transmitter, and the receiver, etc., can also be used separately or alone for the input state. The mobility information may include an indicator whether a RIS is mobile or fixed. Alternatively or in addition, for example, when the RIS is mobile, the mobility information may include information on direction and magnitude of speed of the RIS. Alternatively or in addition, the mobility information may include a current time and a current position of the RIS. Hence, when the RIS moves, it takes possibly a new position at a different time (i.e. new time). Based on the new position and time together with a preceding position and time (i.e. past position and past time), the magnitude and direction of the speed of the RIS can be determined dividing the difference between the new position and the past position by the difference between the new time and the past time. ^ An action is determined by a decision-making function, and it is performed. Said decision- making function corresponds to a policy which is used to map the observation of the current environment to a probability distribution of the action to be taken. This policy may be iteratively learned by the reinforcement learning algorithm. ^ The agent receives a scalar reward. Here, the reward is important, because, in the reward part, all the parameters are simultaneously used. Rewards can be given as follows. The reward will be given for several parameters, which is illustrated by example of the parameter being security. The parameter security may be quantified or characterized by several security measurements, such as secrecy capacity. Moreover, a number of how many attacks have occurred (i.e. detected) in the network may be used to measure security. For example, are there any attacks in the system, such as jamming, eavesdropping, or such? If there is an attack, how many attacks are in the system? Also, in some cases, based on the type of the application type, one may allow some of the attacks to happen, while other attacks may strictly not be wanted. For example, if the parameter reliability is important for the system, one may allow for eavesdropping attacks, but one may not strictly want jamming or spoofing attacks. On the other hand, if parameter secrecy is important, one may not allow eavesdropping, but one may allow for other attacks. If the desired security is provided, the reward is +1. Otherwise, it will be -1. Another example is the parameter RIS load. If the RIS load is high there will be punishment (-1), on the other hand, if it is low there will be a reward. It is noted that whether the RIS load is “high” and “low” or whether “the security is provided” mean that a condition is actually applied on said parameters security and/or RIS load, so as to determine whether they are high or low. For example, such determination may be implemented by use of a threshold and comparing a parameter (i.e. a value of security or RIS load) with a respective threshold. Depending on the reward criteria, a security value higher than a security threshold may be rewarded (e.g. by +1). In turn, a high value of RIS load may be punished / penalized (e.g. by -1). This reward-penalty may be applied for all parameters (i.e. the plurality of RIS parameters may be associated with or assigned to a reward-penalty value). Then, the respective reward-penalty values (e.g. +1 or -1 for each parameter) will be summed up. It is noted that some parameters may be critical for some applications. Like military communication applications, security is critical, In which case the reward and punishment can be higher. In other words, any of the plurality of RIS parameters may be given a priority (e.g. a priority value), based on which the reward-penalty value may be determined. For example, assuming a higher priority value (e.g. +10) indicates higher priority (e.g. priority level 4) than a lower priority value (e.g. +2), then the reward-penalty value may be given to the respective RIS parameter based on the priority level of said RIS parameter. The priority may be used to classify the RIS parameters in terms of priority classes (i.e. groups of RIS parameters). The classification may be performed in that RIS parameters with same priority level belong to a group. Alternatively or in addition, the classification may be performed by defining ranges for the priority level. In this case, RIS parameters having different priority levels but fall within the same priority range belong to a group. For example, the reward-penalty value includes one or more values which are determined according to respective one or more parameters of the plurality of RIS parameters for the two or more RISs. For example, assuming as RIS parameters the RIS load and security for the three RISs 210, 216, and 213 of Fig. 2 forming the communication path, said reward-penalty value may include [(+4,+2)210 ; (+1,+1)216 ; (+1,+5)213]. Thereby, (*,*) carries, for example, the reward-penalty value for the RIS load as first entry, while for the security as second entry. The entries may be arranged in a different manner. However, in case the reward-penalty value includes multiple entries, it may be advantageous that entries are grouped in some ordered manner. In the above example, the reward-penalty value may include the one or more values for RIS load and security for the RISs 210, 216, and 213 as [(+4, +1, +1)load ; (+2, +2, +5)sec]. In this case, the reward-penalty value has a first entry the reward values for the RIS load for the three RISs, and as second entry the respective reward values for security. Note that for some parameters for the reward, the value should be higher than a threshold: for example, throughput, security, and/or channel quality. In other words, some parameters have a minimal reward value. In this manner, their importance for determining the communication path is given a higher weight. However, in some cases, the reward should be less than a threshold: for example, RIS traffic load and/or latency. In other words, for some of other RIS parameters which are known or presumed to affect the communication path less, their respective reward value should have a maximal value. In this manner, an over-weighting of those RIS parameters for determining the communication path is avoided. ^ Information about the reward given for the pair (state , action) is recorded, with the reward indicating the benefit (e.g. as measured by a cost function or reward-penalty function) of taking said action starting from said state. Taking an action transfers the system from a current state to a next / follow-up state. It is noted that different actions may transfer the same current state to different next states. By performing actions, and observing the resulting reward, the policy used to determine the best action for a state can be fine-tuned. Eventually, if sufficient states are observed, an optimal decision policy may be generated, so that an agent performs in that environment in an optimal manner. The term “sufficient” means that, when the applications are working without any problems, or their Bit error rate (BER) performance, secrecy capacities, and throughputs are upper/lower than some respective thresholds, one can say that sufficient states are satisfied. In other words, the sufficiency of the number of observed states is determined by thresholding parameters (e.g. BER, secrecy capacities, and throughputs or the like), i.e. by comparing said parameters with their respective threshold. Note that RISs can be categorized into two groups: active and passive. The active RIS does not only amplify incident signals, but also reconfigures its phases such that the desired signal constructively add at the receiver. This means when the receiver has received from an active RIS an incident signal, the receiver “knows” that the signals constructively added. In turn, when the receiver does not receive the desired signal, the receiver “knows” that the signals did not add constructively. The information on whether or not the receiver received the desired signal is transmitted by the receiver to the active RIS as feedback (feedback information). Based on this feedback information, the RIS may then adjust / reconfigured its phases. In contrast, the capabilities of passive RISs are rather limited when compared with active RISs in terms of adapting or tuning their phases. For passive RIS cases, using simple parameters, such as ToA, ToF, or RSSI may be advantageous, as their use may minimize the complexity of optimizing the communication path. Thereby, the term “simple” means that the respective parameter may be obtained (e.g. calculated) on a low-complexity basis and without exceeding signaling overhead. On the other hand, other parameters such as secrecy capacity or learning traffic load have a higher computational complexity than ToA, ToF, or RSSI. A detailed example of coordinating multiple RISs is given below. Also, Fig.6 represents the basic system model of the RIS network, and Fig.7 represents the RL-algorithm-based solution. Stage 1: Agent Interaction: The RIS network (RIS network means several RISs, which are used to transmit signal Tx to Rx) interaction with the environment can be outlined as follows. The RIS network observes the current state, ^, of the environment and takes an action, ^ , predicated upon the observed state. In other words, the RIS network acts as the agent in Figs.5 and 7, with said RIS network being embedded in a wireless environment. Upon taking said action in the observed state (i.e. the current state), the RIS network takes a new state ^^{^} and receives a reward, ^, for the taken action and receives a new observation for the new state ^^{^} (i.e. next state or follow-up state) from the environment. Once the experience ^^, ^, ^, ^^{^}^ is acquired ,the RIS network trains the RL model using current and past experiences in the second stage. The term “experience” indicates information captured in one learning episode and said information defines the location information of the transmitter and receiver. This location information (^) can be a multipath signature between the transmitter and receiver (e.g. channel information including channel vectors and/or concatenated channel vector), or coordination of the transmitter and receiver for stable/indoor environment. In this example, it is assumed that the one learning episode occurs every coherence block, ^ is the maximum number of episodes, and ^⁽^⁾ denotes the concatenated sampled channel vector at the ^th episode, where ^ = 1, … , ^. Note that ^⁽^⁾ refers to above information (^), but carrying a label ^ indicating the ^-th episode. An coherence block consists of a number of subcarriers and time samples over which the channel response can be approximated as constant and flat-fading. In an exemplary implementation, the channel vector is included in the channel information, with said channel vector including a transmitter channel vector between the transmitter and the plurality of RISs (i.e. transmitter-RISs channel vector) and/or a receiver channel vector between the receiver and the plurality of RISs (i.e. receiver-RISs channel vector). In other words, the channel vector between the transmitter and the receiver is obtained by concatenating the transmitter channel vector and the receiver channel vector. For example, the channel vector in Fig.6 may include transmitter channel vector from Tx to RIS 610 and receiver channel vector from Rx to RIS 610, from Tx to RIS 611 and from Rx to RIS 611, …, from Tx to RIS 61x and from Rx to RIS 61x, and Tx to RIS 61y and from Rx to RIS 61y. Thus, for each of the plurality of RISs, the transmitter-receiver channel vector may be thought of a transmitter-receiver channel vector divided into a channel vector between Tx and one RIS and a channel vector between Rx and the one RIS. The one transmitter-receiver channel vector may then be obtained by joining (i.e. concatenating) the individual Tx-RIS and Rx-RIS channel vectors (i.e. the Tx-RIS and Rx-RIS vectors are sub-channel vectors for said communication). Since there may be more than one RIS, the transmitter-receiver channel vector is composed of a superposition of multiple channel vectors. Since a channel vector (e.g. channel vectors Tx-RIS and Rx-RIS) represents, in case of one RIS, a communication path between the transmitter and the receiver mediated by one RIS, the concatenated channel vector with multiple RISs represents a multi-path between the transmitter and the receiver. Therefore, the concatenated channel vector (i.e. concatenated sampled channel vector) is a part of the multipath signature. The steps performed at Stage 1 are summarized as follows. 1. Estimation of the location information of the transmitter and receiver: The transmitter and receiver transmit two orthogonal uplink (UL) pilots (i.e. pilot signals). The RIS network (e.g. receiver Rx) receives these pilots (from transmitter Tx) and estimates the sampled channel vectors to construct the multipath signature. This means that the RIS network uses the transmitter pilots to determine a transmitter channel vector between the transmitter and the plurality of RISs and/or uses the receiver pilots to determine a receiver channel vector between the receiver and the plurality of RISs. The transmitter channel vector and/or the receiver channel vector are comprised in a channel vector that is in turn included in the channel information. Note that the channel vector between Tx and Rx can be used as location information. The location information can be given in a coordinate format (e.g. GPS, or BS aided location) or such for stable environments. Thus, the sampled channel vectors represent current channel information. 2. Interaction prediction: The location of the transmitter and receiver is used to predict the RISs for the optimal communication path. To account for exploration (i.e., randomly sampling from the action space) besides exploitation (i.e., using prior learning experience), a factor ^ is introduced such that an interaction vector can be randomly chosen out of the codebook ^ with probability ^. Otherwise, the optimum path is predicted from the current network, such as a Q-network that is based on a trained machine-learning model. This means that the network uses as input the current channel information, and obtains the interaction vector from a previous episode. After that, the RISs chosen in the network reflects the transmitted data from the transmitter in that two or more RISs in the communication path are configured to form the communication path between the transmitter and the receiver. 3. Feedback reception: With the communication path formed, the receiver receives data of the communication path and transmits feedback information to the RIS network. The RIS network receives feedback from the receiver indicating one or more of RIS parameters, including the RIS traffic load, security, channel quality, latency, reliability, energy efficiency, throughput, ToA, ToF, ToT, TDoA, and RSSI, ^⁽^⁾, attained by RIS network. Note that here each property is represented by ^⁽^⁾, with the understanding that the label ^⁽^⁾ may be different for different RIS parameters. After that, the feedback can be quantized based on a threshold level ^^{^^} , such that ^_^ ⁽^⁾ = 1 if ^⁽^⁾ > ^^{^^}; otherwise, ^_^(^) = −1, with ^_^(^) referring to a reward-penalty value (also referred shortly as reward). In other words, the RIS network determines a reward-penalty value based on the feedback information. Note that the reward value may take negative or positive values. For example, a positive value indicates reward while a negative value indicates a penalty or punishment. Alternatively, the reward value may be positive or negative only, thereby the reward and penalty then being specified, for example, by an interval of positive or negative values, such as [0,…,Val1] representing an interval of values of reward, while [Val2,…,Val3] representing an interval of values of penalty, with Val3 > Val2 > Val1. Other options for specifying the values for reward and penalty are possible. Moreover, some of the rewards and punishments (i.e. the respective values) can be higher or lower than 1. For example, if the RIS traffic load is important for the system, ^_^ ⁽^⁾ can be equal to 2, if ^⁽^⁾ > ^^{^^}; ^_^ ⁽^⁾ can be -2. Thus, the reward-penalty value may take different values dependent on importance of one or more RIS parameters. This means that, for a RIS parameter, the value of the weight of the reward value may differ from the penalty value (e.g. RIS load: reward value = +4 and penalty value = -1). The values of reward and penalty may take rational and/or irrational values. Moreover, the threshold ^^{^^} may be different for different RIS parameters. Stage 2: Agent Learning: The RIS network leverages the acquired experiences to train the RL model (e.g. the machine-learning model supporting reinforcement learning). The steps performed at Stage 2 are summarized as follows. 1. Constructing a new experience: The new experience acquired is now stored in the experience replay buffer ^ for the training of the Q-network (e.g., deep Q-network). 2. Q-Network training: The Q-network is now trained to minimize the prediction loss. To do so, the stochastic gradient descent algorithm (SGD) can be used. The training operates sequentially using minibatches of experiences from the replay buffer ^ in one training episode. In other words, current and past experiences may be used to train the Q-network within one episode. The Q-network is trained to learn how to map an input state (sampled channel vector) to an output action (interaction vector). Based on the network output (e.g. interaction vector), the communication path is obtained. The Q- network is based on a machine-learning model, and takes as input the channel information that includes the sampled channel vector. Inputted to the Q-network are further the sets of RIS parameters. The Q-network is an example of a machine-learning model, and other networks relying on machine-learning may be suitable for learning how to map an input state (sampled channel vector) to an output action (interaction vector). Further, the machine-learning model supports reinforcement learning (RL). The sequential use of minibatches means that the Q-network (i.e. machine-learning model) is trained in a sequence of episodes, with said episode being a training episode. This means that, in a training episode, the network is trained. Specifically, based on the current channel information that is inputted to the Q-network (i.e. machine-learning model), an interaction vector is obtained as the output of the trained Q-network. Two or more RISs are then configured so as to form the communication path between the transmitter and the receiver. In order to exploit feedback information reflecting whether the communication path may be optimal, the receiver receives data over the communication path, and afterwards transmits the feedback information to the RIS network. This feedback information is received by the control device 300 in Fig.3, which then determines a reward-penalty value based on the received feedback information. New channel information is then obtained by the RIS network. The Q-network is then trained, using the current channel information, the new channel information, the reward- penalty value (e.g. as determined in Step 1), and the interaction vector as current experience. The algorithm is given below to obtain a trained network ^⁽^, ^^|^⁾ for a given current state ^, action ^. The weights ^ of the network that are adapted (updated) during the training. The steps specified within the repeat-until loop are executed until a predetermined optimization goal is reached. repeat RIS network receives two pilots to determine the concatenated channel vector ℎ^{^(}1⁾ between the Tx and Rx. The channel vector ℎ^{^(}1⁾ represents the current state. The channel vector representing the current state is included in the current channel information. As noted before, the concatenated channel vector is a vector obtained by concatenating the transmitter channel vector and the receiver channel vector. The pilots are transmitted by the transmitter and the receiver to the RIS network, respectively. Current state. For episode every episode Stage 1: Agent Interaction Sample ^~Uniform (0,1) if ^ < ^ then Select action Select RIS path (RISs that will be used for transmission), ^(^) ∈ ^ at random. Here, the interaction vector ^(t) is randomly selected from the codebook P. else Select RIS path, ^⁽^⁾ = ^^^^^^_^ ^{^}^⁽^, ^^{^|}^⁾. Hence, based on the current state and action a’, the network (here: Q-network) outputs an interaction vector , based on which a RIS path is selected. As noted before, s refers to a current state, which may be represented by the current channel information that includes transmitter channel vectors and/or receiver channel vectors, obtained by the RIS network upon receiving pilot signals from the transmitter and the receiver respectively. The operation ^⁽^⁾ = ^^^^^^_^ ^{^}^⁽^, ^^{^|}^⁾ means that an action a’ among all available actions is selected that maximizes the output of the Q-network. Here, said output is the interaction vector , based on which a RIS path is selected. RIS network reflects using selected RISs Carry out action. The RIS path includes two or more RISs of the communication path. The selected RISs are then configured to from the communication path, such that the receiver may receive data over this communication path. RIS network receives the feedback ^⁽^⁾ Observe reward. The feedback (i.e. feedback information) is transmitted to the RIS network. Observing reward means that the RIS network determines, based on the received feedback, a reward-penalty value for the sets of RIS parameters, i.e. each RIS parameter in a set is given a value for reward or penalty, as described above. RIS network receives two pilots to estimate ℎ^{^}(^ + 1) ^ Next state. Then, with the RISs determined based on the output interaction vector, the transmitter and the receiver transmit pilots to the RIS network, so as to obtain new channel information, including a new channel vector ℎ^{^}(^ + 1). The argument t+1 reflects that the channel vector ℎ^{^(}^ + 1⁾ represents the next state different from the current state ℎ^{^(}^⁾. In other words, ℎ^{^(}^⁾ is obtained from an experience preceding the (next or following) experience where ℎ^{^(}^ + 1⁾ is obtained. The next state may be also refer to a follow-up state. Stage 2: Agent Learning The above assignment means that the current state s is represented by the current channel information (e.g. channel vector), the next / new / follow-up state is represented by the new channel information (e.g. channel vector), the action a is represented by the interaction vector, and r refers to the reward-penalty value when the state changes from state s to s’ when the action a is taken. The interaction vector (i.e. action) is the best path based on the current channel vector and the past experience. Store the experience ^{^}^, ^, ^, ^^{^^} in ^ (i.e. replay buffer). Minibatch experiences from ^ for training. This means that the machine-learning model (e.g. Q-network) is trained using a subset of experiences from buffer D. This means that, in a training episode, for example, the subset may have 100 experiences taken from 100,000 experiences stored in the buffer. feedforward ^ to calculate a reward-penalty value ^^{^(}^⁾ ⇐ ^⁽^, ^^|^⁾∀^. Here, the current state a of an experience from the minibatch experiences is provided as input to the Q-network, along with an action a among all actions that can be taken from current state s. For each pair (s,a), the Q-network provides as output an interaction vector, and a respective reward-penalty value ^^{^(}^⁾ is calculated. For example, assume that for an experience associated with a current state s there are 5 actions a. Then, 5 reward-penalty values are calculated. feedforward ^^{^} to calculate ^ ⇐ max^{^}^ ^(^^{^}, ^^{^}|^). Here, the maximum interaction vector is determined from the Q-network that is provided as input the next state s’ and all the actions a’ that can be taken from said new state s’. In other words, for given new state s’, the Q-network is provided an action a’ as input and provides an interaction vector as output. This calculation is done for all actions a’ that can be taken from new state s’. Among the calculated interaction vectors (i.e. the number of interaction vectors corresponds to the number of actions a’), the maximum is then determined. and calculate ^^∗ ⇐ ^^^max^{^}^ ^⁽^^{^}, ^^{^|}^⁾. Here, the Q-network is provided as input the next state s’ and an action a’. The operation argmax^a’ means finding the action a’ that maximizes the output of the network, which is the interaction vector. Construct the target vector, ^^{^(}^⁾ which may be an achievable rate: Here, one a is selected from a’ of the codebook P so as to find the best path for the system with achievable rate. Perform stochastic gradient descent (SGD) on the mean-squared-error to find we ^∗ ights ^ . Note that the mean-squared error (MSE) is an example of a loss function, with the SGD algorithm applied on the MSE hence finding the minimum of the loss. This represents then the optimal solution with the loss being minimized. As noted before, said optimal solution (i.e. optimal communication path) may not yet be the best (global) optimum that can be achieved. Rather, the optimal solution represents the optimum for this set of RIS parameters, channel information, etc. The best achievable optimum is obtained when a (predefined) terminal goal is approached. Update network weights ^⁽^⁾ ⇐ ^^∗ Decrease ^ gradually. Please note that the decrease of ^ is motived in that, at the beginning of the training, there is not yet a sufficient number of experiences obtained from interactions vectors ^(^) = ^^^^^^_^^^(^, ^^{^}|^). Hence, the likelihood for selecting an interaction vector from a codebook P (e.g. predefined) may be larger at the beginning of the training than at the end and/or during the progressed training. Therefore, with increasing training progress, more interaction vectors are obtained and hence more experiences are available for the training. As a result, the likelihood ^ can be gradually reduced with increased training progress. In this case, it becomes more likely that the interaction vector is determined from = ^^^^^^_^^^⁽^, ^^{^|}^⁾, but less or very unlikely being randomly selected from the codebook P. ^ ⇐ Assign next state to current state. until reaching a terminal goal Hence, in each episodes t = 1 to T, the machine-learning model is trained, so that the episode is also referred to as training episode. An illustration of above RL-based RIS coordination is shown in Fig.7, where a plurality of RISs 710 to 715, …, 71x, …, 71y may mediate communication between the transmitter Tx and the receiver Rx. As shown, the RISs along with a reinforcement learning module represent the agent of Fig.5. Between the RISs, there may be various kinds of objects impacting the communication between Tx and Rx. Hence, an optimal communication path are to be found that includes two or more RISs among the plurality of RISs. As discussed above, Tx and Rx send (transmit) pilot signals to the RIS network formed by the RISs 710 to 715, …, 71x, …, 71y, where one of the RISs determines channel information, including channel vectors such as transmitter channel vector and/or receiver channel vector. Based on the channel information along with sets of RIS parameters, the RIS network determines for the sets of RIS parameters a reward-penalty value and an interaction vector to determine two or more RISs of the communication path. The two or more RISs are configured to for the communication path. The Tx and Rx transmit again pilot signals to the RIS network over the RISs forming the communication path, so that the RIS network determines new channel information corresponding to the next state. In other words, the receiver Rx provides feedback information to the RIS network for the current state and the next state. Since the RIS network acts as agent, it uses the information of the current state and the next state along with the reward (i.e. reward-penalty value) to determine, by training of a machine-learning model (e.g. RL learning), an interaction vector. The RIS coordination may be performed by one of the RISs 710 to 715, …, 71x, …, 71y in Fig.7, comprising communication module 300 and processing circuitry 320 of Fig.3. In one exemplary implementation, the control device 300 of one RIS may have separate modules for training the machine learning model (i.e. module 330) and for executing the statistical algorithm (i.e. module 340). These modules 330 and 340 bi-directionally communicate with the processing circuitry 320 controlling the particular processing for the learning and processing the statistical algorithm. Alternatively, any of modules 330 and/or 340 may be part of the processing circuitry. Fig.4 shows a flowchart for performing RIS coordination, where in step 410, channel information is obtained, with said channel information relating to a channel between the transmitter Tx and the receiver Rx. For each RIS, a set of RIS parameters is obtained in step S420, with the set including a plurality of parameters that are related o said RIS. In an exemplary implementation, the steps 410 and 420 may be executed by communication modules 310 of control device 300 in Fig.3. Based on the sets of RIS parameters and on the channel information, a communication path is determined in step 430, which may be executed by processing circuitry 320 of control device 300 in Fig. 3. The flowchart in Fig. 4 further shows steps 432 and 434 which may be subsumed in step 430 and which refer to steps for applying the trained machine-learning model and applying the statistical learning algorithm, respectively. In the example case, where the control device 300 in Fig.3 is configured with separated modules 330 and 340, the processing circuitry 320 may then control modules 330 and 340 such that step 432 is executed by module 330 and step 434 is executed by module 340, respectively. It is noted that this is a mere implementation example for which module of the control device executes which step of the flowchart in Fig.4. Note that each RIS can estimate the reward and handle it accordingly. Also, a control entity (e.g. a controller) can calculate the optimum path with knowledge of all. This means that each RIS can estimate the reward and handle it accordingly, or a controller gets the reward from other RISs and decides on the optimum paths for each RIS. This control entity can be a BS or smart RIS with powerful processing capability. The RIS coordination has been discussed with applying a trained machine-learning model to determine the communication path. In particular, with the machine- learning model supporting reinforcement learning, the communication path is determined simultaneously with the training of the machine-learning model. Alternatively, the communication path may be determined by applying a statistical algorithm. The above discussed RIS coordination may be applied alternatively to other scenarios, without departing from the above approach according to the present disclosure, including: ^ Multiple RIS coordination framework. ^ Doing RIS coordination with existing signals in the environment. The existing signals may include reflected signals, which are generated from other devices. For example, when a device generates a signal, the RIS can reflect it and use it for coordination purposes. ^ Using a base station (BS) for the coordination. This approach is illustrated in Fig.8. Here, a BS does the coordination between the RISs 810 to 820. For the coordination, the parameters discussed before (i.e. the various RIS parameters) can be used. Note that, instead of the BS, a master device can be selected from the RISs, which will have extra capability to process data for the coordination (e.g. processing circuitry for RISs coordination). In other words, a RIS performs the coordination of the RISs, having a configuration as the one shown in Fig.3, for example. Such RIS selected among the RISs 810 to 820 may be RIS 810 that is closest to the transmitter Tx. Alternatively, RIS 812 may be used that is closest to the BS. ^ More than one BS can be used for the coordination. Therefore, location information of BSs and RIS can be included to optimization. Hence, also other RISs may be configured with processing circuitry to perform RIS coordination. In this case, the other RISs may deactivate (e.g. by respective internal signaling) their processing circuitry and/or modules needed for RIS coordination. Still having one or more RISs in addition to the above selected RIS may allow switching the RIS coordination to another RIS in case said selected RIS may fail. In this way, the RIS coordination may be become more robust against single-RIS failure. ^ RIS can be more intelligent in the future and make coordination using their own signals and processing units. In a multi-RIS scenario, some of the RISs can have more processing capability so as to perform RIS coordination, while other RISs are simple in that their processing circuitry make perform limited tasks for RIS coordination that are less complex. For example, RIS systems may not generate a signal, but rather they use reflected signals to reflect them intelligently to the receivers. However, just for the case of coordination, more intelligent RISs can be used, which are capable of generating their own signals. ^ Some RISs are very simple. For them, a sensor or this kind of device can be used to calculate basic parameters such as ToA, ToF, and/or RSSI. ^ The coordination may not be limited to selecting the best path, but may be also used for addressing diverse requirements of 5G and beyond networks, such as mobility management, reliability and latency, energy efficiency, security, etc.. ^ For the weighting of reward (i.e. using different reward values), different levels can be used. For example, there can be 5 scores for security (i.e. levels of security, which represent the security levels. For example, -2, -1, 0, 1, and 2. In this example, -2 and -1 are the punishments, while 1 and 2 are the rewards. In other words, the values indicate besides the type of the reward-penalty (e.g. positive value means reward, and negative value means penalty) via their sign, as well as a particular level associated with another quantity. In this example, the other quantity is security having various levels of security. This has been also detailed further above. ^ The RIS system (i.e. the multiple RIS) can coordinate with other RIS networks. Fig. 9 shows an example of two RIS systems, in which case the RISs of both RIS systems 1 and 2 would be coordinated. As shown, system 1 entails RISs 910 to 919 and system 2 entails RISs 920 to 929. RISs 910, 916, and 913 of system 1 form a communication path between the transmitter and the receiver. Likewise, RISs 920, 926, and 923 of system 2 form a communication path between the transmitter and the receiver. The transmitter and the receiver in system 1 may be different from the transmitter and the receiver in system 2, Such scenario may be realized for RIS system 1 being located in a first wireless cell (e.g. LTE cell) while the RIS system 2 being located in a second cell (e.g. LTE cell), with the first and second cell being different. For example, the first and second cell may be an LTE cell, where the first LTE cell neighbors the second LTE cell. Alternatively, the transmitter and the receiver in system 1 and 2 be the same, but use in parallel different RISs systems. In either case, the RIS coordination may be performed by one of the RISs of the RISs of system 1 and 2. For example, RIS 923 of system 2 may be selected for the coordination of the RISs of system 1 and 2. Coordinated reconfigurable intelligent surfaces (CRIS) can address the diverse requirements of 5G and beyond networks, such as mobility management, reliability and latency, energy efficiency, security, and the like. Also, when the optimum path or resource allocation is made, not only a single transmission may be considered, but the overall system performance should be considered. This is especially the case for RIS capacity. This is because a RIS may be selected as being part of the optimal RISs (i.e. optimal path) for different transmissions, and said RIS may be used for these different transmissions. However, in this case, the network load of this RIS may be likely rather high compared to another RIS which is optimal for only one transmission. Thus, the different transmissions of said optimal RIS may affect its performance. Therefore, if the RIS is trying to be used for several transmissions, it can be possible to use other RIS with the coordination. In the present disclosure this is inherently accounted for in the manner in which the optimum path is determined, since the network load of the system is considered. The RIS also can be used with different RISs in other systems. Therefore, with the coordination of the RISs, the overall system and even other network systems may be optimized. For RIS coordination, ML- based algorithms are advantageous, because they allow considering simultaneously several parameters and networks. Especially an RL-based strategy may be promising, because the importance of the parameters can be determined manually with the rewards and punishments. Also, in the RL algorithm, the model is trained and tested simultaneously. Note that it is difficult to realize RIS coordination with other algorithms, such as supervised learning-based classification. In this algorithm, there is a need to have classes for each possible path which is an unrealistic approach and the training complexity of it is very high. The need to have classes is reasoned in that, in a supervised learning algorithm, a labelled dataset is needed, which is difficult to realize in this scenario of RIS coordination. This is because, in RIS coordination, there are several paths that a signal can go over. For example, based on the coordination, the signal can use RIS210, RIS213, RIS215, RIS216 of Fig.2, and reach the receiver. Or it can use RIS212, RIS214, RIS216, and reach the receiver or such. For the case of the supervised algorithm, all of these scenarios should be labelled, and there should be several examples of these scenarios in the training dataset to realize the algorithm with a supervised learning algorithm. In some embodiments, the processing circuitry performing the functions described herein may be integrated within an integrated circuit on a single chip. Moreover, the controller 300 in Fig.3 may include memory (not shown in Fig.3) which may store a plurality of firmware or software modules, which implement some embodiments of the present disclosure. The memory may be read from by the processing circuitry 320. Thereby, the processing circuitry may be configured to carry out the firmware/software implementing the embodiments. The processing circuitry 320 may include one or more processors, which, in operation, perform the above method steps for determining a communication path. The transmitter (transceiver, transmitting device) shown in Figs.2, and 5 to 8, may be, for example a wireless transceiver. Likewise, the receiver (receiving device) shown in Figs.2, and 5 to 8 may be, for example a wireless receiver. The present disclosure may be used in multiple RIS-aided wireless networks. It can be used in cellular and Wireless Fidelity (Wi-Fi) networks. The present disclosure has the potential to be an essential standard in 6G technologies. In 2021, the European Telecommunications Standards Institute (ETSI) launched a new group on Reconfigurable Intelligent Surfaces. The present disclosure can be essential for that group. The present disclosure has a potential to be essential standard in 6G technologies. In summary, the present disclosure provides for techniques and methods to coordinate multiple reconfigurable intelligent surfaces (RISs), so as to determine a communication path between a transmitter and a receiver communicating via a channel. In order to determine the communication path, channel information that relates the channel between the transmitter and the receiver is obtained. Further, for each RIS of the multiple RISs, a set of RIS parameters is obtained, where the set includes multiple parameters related to said RIS. The sets of RIS parameters and the channel information are then used to determine the communication path that includes two or more RISs. Coordinating multiple RISs in that their multiple RIS parameters are jointly used, enables determining a more robust communication path, and hence makes the communication between a transmitter and a receiver more stable. This because the RISs are embedded in a wireless environment and hence interact, thereby impacting the communication path (i.e. the channel) used by the transmitter and the receiver. Thus, a communication path is determined in account of RIS interaction, as realized by RIS coordination and exploiting their RIS parameters. Implementations in software and hardware The methodologies described herein may be implemented by various means depending upon the application. For example, these methodologies may be implemented in hardware, operation system, firmware, software, or any combination of two or all of them. For a hardware implementation, any processing circuitry may be used, which may include one or more processors. For example, the hardware may include one or more of application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, any electronic devices, or other electronic circuitry units or elements designed to perform the functions described above. If implemented as program code, the functions performed by the control device 300 may be stored as one or more instructions or code on a non-transitory computer readable storage medium. The computer-readable media includes physical computer storage media, which may be any available medium that can be accessed by the computer, or, in general by the processing circuitry 320. Such computer-readable media may comprise RAM, ROM, EEPROM, optical disk storage, magnetic disk storage, semiconductor storage, or other storage devices. Some particular and non-limiting examples include compact disc (CD), CD-ROM, laser disc, optical disc, digital versatile disc (DVD), Blu-ray (BD) disc or the like. Combinations of different storage media are also possible – in other words, distributed and heterogeneous storage may be employed. For example, the program code may cause the processing circuitry 320 (e.g. including one or more processors) to operate as a special purpose computer programmed to perform the techniques disclosed herein. The embodiments and exemplary implementations mentioned above show some non-limiting examples. It is understood that various modifications may be made without departing from the claimed subject matter. For example, modifications may be made to adapt the examples to new systems and scenarios without departing from the central concept described herein. In particular, the above embodiments and exemplary implementations are multiple-input multiple-output (MIMO) compatible and can be applied to all MCSs. Selected embodiments and examples According to an aspect, a control device is provided for coordination of a plurality of reconfigurable intelligent surfaces, RISs, the control device comprising: a communication module configured to: obtain channel information related to a channel between a transmitter and a receiver; and obtain, for each RIS of the plurality of RISs, a set of RIS parameters including a plurality of parameters related to said RIS, and processing circuitry configured to: determine a communication path between the transmitter and the receiver based on the channel information and based on the sets of RIS parameters of the plurality of RISs, wherein the communication path includes two or more RISs of the plurality of RISs. In some exemplary implementations, the processing circuitry is configured to control the communication module to provide path information regarding the communication path to the two or more RISs included in the communication path. For example, the path information provided to a specific RIS of the two or more RISs includes: identification of a RIS preceding said specific RIS in the communication path; and/or identification of a RIS following said specific RIS in the communication path. For example, the plurality of RIS parameters includes one or more of traffic load; security; channel quality; latency; reliability; energy efficiency; throughput; time of arrival, ToA; time of flight, ToF; time of transmission, ToT; time difference of arrival, TDoA; and received signal strength indicator, RSSI. Moreover, the determining of the communication path is performed by applying a statistical algorithm or by applying a trained machine-learning model. In an implementation, the determining of the communication path includes: inputting, to the trained machine-learning model, the channel information and the sets of RIS parameters of the plurality of RISs; and obtaining, according to the output of said trained machine-learning model, the communication path. According to an implementation example, the machine-learning model supports reinforcement learning in a training episode and includes: obtaining an interaction vector as the output of said trained machine-learning model from a previous episode, based on a current channel information inputted to the machine-learning model; configuring the two or more RISs in the communication path to form the communication path for the transmitter and the receiver; receiving feedback information from the receiver after the receiver received data over the communication path; determining a reward-penalty value based on the feedback information; obtaining next channel information; and using the current channel information, the new channel information, the reward- penalty value, and the interaction vector as current experience to train the machine-learning model. For example, the reward-penalty value includes one or more values determined according to respective one or more parameters of the plurality of RIS parameters for the two or more RISs. Further, the channel information includes location information of the receiver and/or location information of the transmitter. In a first implementation, the communication module is configured to obtain the channel information from the receiver, wherein the channel information includes a value of a signal strength measurement of a reference signal. In a further implementation, the channel information includes a channel vector comprising: a transmitter channel vector between the transmitter and the plurality or RISs; and/or a receiver channel vector between the receiver and the plurality of RISs. For example, the communication module is configured to provide to the transmitter identification of a RIS of the two or more RISs that is closest to the transmitter and pertaining to said communication path. Moreover, the communication module is configured to provide to the receiver identification of a RIS of the two or more RISs that is closest to the receiver and pertaining to said communication path. In a further example, the control device is an access point or a RIS among the plurality of RISs. Moreover, the receiver and/or the transmitter is a wireless station and/or a RIS. According to an aspect, a method is provided for coordination of a plurality of reconfigurable intelligent surfaces, RISs, the method comprising: obtaining channel information related to a channel between a transmitter and a receiver; obtaining, for each RIS of the plurality of RISs, a set of RIS parameters including a plurality of parameters related to said RIS, and determining a communication path between the transmitter and the receiver based on the channel information and based on the sets of RIS parameters of the plurality of RISs, wherein the communication path includes two or more RISs of the plurality of RISs. The examples and exemplary implementations described above for the control device apply in the same manner to the above method. In particular, the processing circuitry may be further configured to perform method steps corresponding to the processing of the control device of one or more of the above-described embodiments and exemplary implementations. Still further, provided is a computer program stored on a non-transitory and computer-readable medium, wherein the computer program includes instructions which when executed on one or more processors or by a processing circuitry perform steps of any of the above-mentioned methods. According to some embodiments, the processing circuitry and/or the communication module is embedded in an integrated circuit, IC. Although the disclosed subject matter has been described in detail for the purpose of illustration based on what is currently considered to be the most practical and preferred embodiments, it is to be understood that such detail is solely for that purpose and that the disclosed subject matter is not limited to the disclosed embodiments, but, on the contrary, is intended to cover modifications and equivalent arrangements that are within the spirit and scope of the appended claims. For example, it is to be understood that the presently disclosed subject matter contemplates that, to the extent possible, one or more features of any embodiment can be combined with one or more features of any other embodiment.

Claims

CLAIMS 1. A control device for coordination of a plurality of reconfigurable intelligent surfaces, RISs, the control device comprising: a communication module configured to: ^ obtain channel information related to a channel between a transmitter and a receiver; and ^ obtain, for each RIS of the plurality of RISs, a set of RIS parameters including a plurality of parameters related to said RIS, and processing circuitry configured to: ^ determine a communication path between the transmitter and the receiver based on the channel information and based on the sets of RIS parameters of the plurality of RISs, wherein the communication path includes two or more RISs of the plurality of RISs.

2. The control device according to claim 1, wherein the processing circuitry is configured to control the communication module to provide path information regarding the communication path to the two or more RISs included in the communication path.

3. The control device according to claim 2, wherein the path information provided to a specific RIS of the two or more RISs includes: identification of a RIS preceding said specific RIS in the communication path; and/or identification of a RIS following said specific RIS in the communication path.

4. The control device according to any of claims 1 to 3, wherein the plurality of RIS parameters includes one or more of traffic load; security; channel quality; latency; reliability; energy efficiency; throughput; time of arrival, ToA; time of flight, ToF; time of transmission, ToT; time difference of arrival, TDoA; and received signal strength indicator, RSSI.

5. The control device according to any of claims 1 to 4, wherein the determining of the communication path is performed by applying a statistical algorithm or by applying a trained machine-learning model.

6. The control device according to claim 5, wherein the determining of the communication path includes: inputting, to the trained machine-learning model, the channel information and the sets of RIS parameters of the plurality of RISs; and obtaining, according to the output of said trained machine-learning model, the communication path.

7. The control device according to claim 6, wherein the machine-learning model supports reinforcement learning in a training episode and includes: obtaining an interaction vector as the output of said trained machine-learning model from a previous episode, based on a current channel information inputted to the machine- learning model; configuring the two or more RISs in the communication path to form the communication path for the transmitter and the receiver; receiving feedback information from the receiver after the receiver received data over the communication path; determining a reward-penalty value based on the feedback information; obtaining new channel information; and using the current channel information, the new channel information, the reward-penalty value, and the interaction vector as current experience to train the machine-learning model.

8. The control device accord to any of claims 1 to 7, wherein the reward-penalty value includes one or more values determined according to respective one or more parameters of the plurality of RIS parameters for the two or more RISs.

9. The control device according to any of claims 1 to 8, wherein the channel information includes location information of the receiver and/or location information of the transmitter.

10. The control device according to any of claims 1 to 9, wherein the communication module is configured to obtain the channel information from the receiver, wherein the channel information includes a value of a signal strength measurement of a reference signal.

11. The control device according to any of claims 1 to 10, wherein the channel information includes a channel vector comprising: a transmitter channel vector between the transmitter and the plurality or RISs; and/or a receiver channel vector between the receiver and the plurality of RISs.

12. The control device according to any of claims 1 to 11, wherein the communication module is configured to provide to the transmitter identification of a RIS of the two or more RISs that is closest to the transmitter and pertaining to said communication path.

13. The control device according to any of claims 1 to 12, wherein the communication module is configured to provide to the receiver identification of a RIS of the two or more RISs that is closest to the receiver and pertaining to said communication path.

14. The control device according to any of claims 1 to 13, wherein the control device is an access point or a RIS among the plurality of RISs.

15. The control device according to any of claims 1 to 14, wherein the receiver and/or the transmitter is a wireless station and/or a RIS.

16. A method for coordination of a plurality of reconfigurable intelligent surfaces, RISs, the method comprising: obtaining channel information related to a channel between a transmitter and a receiver; obtaining, for each RIS of the plurality of RISs, a set of RIS parameters including a plurality of parameters related to said RIS, and determining a communication path between the transmitter and the receiver based on the channel information and based on the sets of RIS parameters of the plurality of RISs, wherein the communication path includes two or more RISs of the plurality of RISs.

17. The method according to claim 16 including providing path information regarding the communication path to the two or more RISs included in the communication path.

18. The method according to claim 17, wherein the path information provided to a specific RIS of the two or more RISs includes: identification of a RIS preceding said specific RIS in the communication path; and/or identification of a RIS following said specific RIS in the communication path.

19. The method according to any of claims 16 to 18, wherein the plurality of RIS parameters includes one or more of traffic load; security; channel quality; latency; reliability; energy efficiency; throughput; time of arrival, ToA; time of flight, ToF; time of transmission, ToT; time difference of arrival, TDoA; and received signal strength indicator, RSSI.

20. The method according to any of claims 16 to 19, wherein the determining of the communication path is performed by applying a statistical algorithm or by applying a trained machine-learning model.

21. The method according to claim 20, wherein the determining of the communication path includes: inputting, to the trained machine-learning model, the channel information and the sets of RIS parameters of the plurality of RISs; and obtaining, according to the output of said trained machine-learning model, the communication path.

22. The method according to claim 21, wherein the machine-learning model supports reinforcement learning in a training episode and includes: obtaining an interaction vector as the output of said trained machine-learning model from a previous episode, based on a current channel information inputted to the machine- learning model; configuring the two or more RISs in the communication path to form the communication path for the transmitter and the receiver; receiving feedback information from the receiver after the receiver received data over the communication path; determining a reward-penalty value based on the feedback information; obtaining new channel information; and using the current channel information, the new channel information, the reward-penalty value, and the interaction vector as current experience to train the machine-learning model.

23. The method accord to any of claims 16 to 22, wherein the reward-penalty value includes one or more values determined according to respective one or more parameters of the plurality of RIS parameters for the two or more RISs.

24. The method according to any of claims 16 to 23, wherein the channel information includes location information of the receiver and/or location information of the transmitter.

25. The method according to any of claims 16 to 24 including: obtaining the channel information from the receiver, wherein the channel information includes a value of a signal strength measurement of a reference signal.

26. The method according to any of claims 16 to 25, wherein the channel information includes a channel vector comprising: a transmitter channel vector between the transmitter and the plurality of RISs; and/or a receiver channel vector between the receiver and the plurality of RISs.

27. The method according to any of claims 16 to 26 including providing to the transmitter identification of a RIS of the two or more RISs that is closest to the transmitter and pertaining to said communication path.

28. The method according to any of claims 16 to 27 including providing to the receiver identification of a RIS of the two or more RISs that is closest to the receiver and pertaining to said communication path.

29. The method according to any of claims 16 to 28, wherein the method is executed by a control device being an access point or a RIS among the plurality of RISs.

30. The method according to any of claims 16 to 29, wherein the receiver and/or the transmitter is a wireless station and/or a RIS.

31. A computer program stored on a non-transitory and computer readable medium, wherein the computer program includes instructions which when executed on one or more processors perform the method according to any of claims 16 to 30.