Computerized systems and methods for enhancing MIMO hybrid-beamforming under interference

Abstract

Provided herein are computer implemented methods and systems for optimal MIMO hybrid-beamforming under static and / or dynamic interferences to optimize and enhance a received wireless signal transmitted from a transmitter.

Description

[0001] COMPUTERIZED SYSTEMS AND METHODS FOR ENHANCING MIMO HYBRIDBEAMFORMING UNDER INTERFERENCE

[0002] FIELD OF THE INVENTION

[0003] The disclosure relates to computerized systems and methods for optimal MIMO Hybridbeamforming under static and / or dynamic interference to optimize / enhance a received wireless signal transmitted from a transmitter.

[0004] BACKGROUND

[0005] Rapid, effective, and accurate decoding of simultaneously transmitted ultra-rate symbols in advanced Multiple-Input-Multiple-Output (MIMO) wireless communication systems under the complicated spectrum of various interference phenomena is one of the most challenging tasks in the last decades. The wide range of interference includes destructive effects such as interference signals from general users, smart jamming devices, random scattering objects in the channels, and substantial multipath effects.

[0006] Most of the advanced analog-digital Soft-Interference Cancellation (SIC) algorithms in MIMO setup require perfect or accurate Channel-State-Information (CSI) or, at least, Channel-Statistical-Distribution Information (CDI). Another complex and necessary requirement is the implementation of physical feedback between the receiver and the transmitter to share the CSI / CDI. Applying exact decoding requirements of ultra-rate symbols under the challenging interference assumptions mentioned above and the premise of frequency -fast-fading is almost impossible with the help of conventional SIC algorithms or classical MIMO -beamforming techniques.

[0007] There are generally two approaches to dealing with disturbances and interference signals effects in the space under CSI uncertainty while utilizing an ML tool: The SicNet and Massive MIMO-beamforming based on ML tools.

[0008] The SicNet approach is a deep learning-aided SIC detector termed SICNet, which replaces the interference cancellation blocks of classical SIC by DNN and learns the detection rule in the context of Non Orthogonal-Multiple-Access (NOMA) systems. This method is based on an iterative SIC scheme and learns to carry out joint detection from a limited set of training samples. These approaches do not assume dealing with an environment of fast-selective frequency channels,fast-time-variant channels or dealing with an environment dynamic interference resulting from significantly stronger signals from neighboring user equipment or significantly stronger signals from intelligent jamming techniques concerning the desired received signal in a Stand-Alone (SA)-unlicensed system. The direct meaning of this strict assumption is the appearance of interferences that cause damage to leading parameters in the decoding technique (for example, interferences that cause the same provider to be receptive to all users without distinguishing between several providers among the users, i.e., a complete collapse of NOMA techniques). Thus, a foremost hurdle of this approach is in breaking the process of learning the legality-decoding, both in the learning-testing-validation process and in the online real-time process.

[0009] The second approach to dealing with strong interference signals is the approach of high-resolution beamforming techniques based on ML. The millimeter (mm)Wave massive MEMO can integrate a wide range of technologies, including multi-layering of beams based on channel interference cancellation and capacity enhancement simultaneously, creating significant energy and spectrum efficiency and classic intelligent beam design, including strategies based on ML techniques. The main disadvantage of this approach is that the beams, even if the design beam and the nulls are correct relative to the user and the interference, those beams are dispersed under rich scatters and rich interference environment. Another disadvantage of a massive MLMO-beamforming approach that eliminates channel interference and destructive effects is the heavy complexity of beam planning and direction of arrival (DOA) estimation. The main challenge is that building the beam shape for the user and combining blank null steering against main lobe interference is difficult in a dynamic spectrum that produces a massive interference pattern change since the central beams in both the transmitter and the receiver are scattered due to hitting the medium that combines scattering and interference. Therefore, it is impossible to significantly increase the signal-to-noise ratio (SNR) level in space in definitive non-flexible directions and to produce several spatially independent beams.

[0010] There is therefore a need for computerized systems and methods for enhanced MIMO Hybrid-beamforming under static and / or dynamic interference(s), to enhance and optimize a received wireless signal transmitted from a transmitter.SUMMARY

[0011] According to some embodiments, provided herein are computerized systems and methods for enhanced MIMO Hybrid-beamforming under static and / or dynamic interference(s) to enhance and optimize a received wireless signal transmitted from a transmitter.

[0012] According to some embodiments, there are provided herein method and system for MIMO-hybrid-beamforming interference offset, based on machine learning and classification tools (for example, artificial neural networks, such as radial basis functions (RBFs), including Radial Basis Function Neural Networks ((RBFNN)) and classification algorithms (such as K-means and Recursive Least Squares (RLS) algorithms). To this aim, a new architecture of a MIMO-hybrid beamforming receiver is provided that can be utilized under various conditions, including, for example, dynamic interference, scatters, and multi-path effects, such as delay-spread and linear or nonlinear channels.

[0013] In some embodiments, the disclosed methods and systems advantageously allow self-computational feedback at the receiver side. Moreover, they do not require any information on the Channel-State-Information (CSI), Channel-Statistical-Distribution (CDI), or any feedback between the transmitter and the receiver.

[0014] According to some embodiments, the disclosed methods and systems can advantageously be used to approximate real-time, online wireless communication while combining analog-digital matrices that can optimize the receiver performance and the decoding process in wireless MIMO channels under challenging environments.

[0015] According to some embodiments, advantageously, the disclosed systems and methods not only improve the performance of MLMO-beamforming systems but also offer a solution to conflicts that cannot be resolved by conventional techniques, including, for example, optimal balancing between the interference cancellation and the increment of the channel’s capacity.

[0016] According to some embodiments, the systems and methods disclosed herein provide various advantages intertwined with each other, which produce technological superiority in the field of wireless communication, including solving unsolved engineering problems in the area of interference and in the field of the dynamic spectrum as compared to classical methods; capable of being integrated into the standardization of international communication related to base stationarchitecture, while transforming the classic base stations into autonomous cognitive base stations based on machine learning; and formulation of efficient solutions for dealing with communication interference in the wireless field, that is, offline and online service, and vice versa.

[0017] According to some embodiments, there is provided a computer-based method for optimal MIMO Hybrid-beamforming under static and / or dynamic interference to optimize a received wireless signal transmitted from a transmitter, the method includes:

[0018] a. receiving, by a receiver unit, an initial received wireless signal (Y), transmitted from a transmitter (transmitting a signal (S)), and an estimated combining matrix (estimated Uest);

[0019] b. calculating an initial interference matrix Wmi and a weight matrix Bini, based on the initial Y and the estimated Uest;

[0020] c. calculating, using a machine learning (ML) algorithm, trained to determine a nonlinear function AH, which correlates an interference matrix (W) and a combining matrix (U), and based on the calculated initial Wini and Bini, an initial optimal Uiopt; d. calculating, using the trained ML algorithm, based on one or more iterations of the determined non-linear function, a final W (Wfin) and a final B (Bfin), to thereby determine a final optimal Ufopt, , wherein for each step, the combining matrix is corrected according to a closed-form formula for the optimal U; and

[0021] e. adjusting beamforming of the receiver unit, based on Ufopt, thereby optimizing the received wireless signal by the receiver unit.

[0022] According to some embodiments, the algorithm is executed in the receiver unit.

[0023] According to some embodiments, the method is executed in real-time.

[0024] According to some embodiments, the wireless signal may be a LTE, 5G-NR, mobile 6th generation, Wi-Fi 6-9, Bluetooth, satellite communication signal, GPS, GLONASS, Galileo signal, or any combinations thereof.

[0025] According to some embodiments, wireless signal frequency may be below 6GHz, or in the range of about 6-100GHz.According to some embodiments, the optimization of the received wireless signal is facilitated at the receiver side only.

[0026] According to some embodiments, the optimization of the received wireless signal does not require feedback, a closed-loop-MIMO and / or operation of Physical Upload Share Channels (PUS CH) between the transmitter and the receiver.

[0027] According to some embodiments, the receiver unit is a base station, an e / g-nodeB component, a relay, a small cell, a point-to-point transceiver, or any combinations thereof.

[0028] According to some embodiments, the receiver unit is an end unit.

[0029] According to some embodiments, the ML algorithm is trained offline, by classifying a collected database comprising a plurality of data patterns, to obtain one or more clusters, wherein each point in a cluster comprises a static interference matrix (Wij) and a control combining matrix (Uj) data set points, having a similar joint influence on the received signal (Yij), wherein the joint influence effect is determined based on a weight matrix Bij; and

[0030] determining, using an ML algorithm, a non-linear function model for the AH, correlating the static interference matrix (W) and the combining matrix (U) and influence thereof on the received signal.

[0031] According to some embodiments, the data patterns may be created from various interfering areas, said data patterns includes channel MLMO-wireless interference matrices (AHi, 1=1,..., L), static interference matrices (Wi), and control combining matrices (Ui).

[0032] According to some embodiments, the interference may include: interference signals from general users, interference signals from smart jamming devices, random scattering objects in the wireless channels, substantial multipath effects, active jammers, passive jammers, or any combinations thereof.

[0033] According to some embodiments, there is provided a system for optimal MLM0-beamforming under interference, to optimize a received wireless signal, the system includes a processor configured to execute the method disclosed herein.

[0034] According to some embodiments, the system may further include a receiver unit functionally associated with the processor.According to some embodiments, there is provided an electronic device having one or more processors; and memory coupled to the one or more processors, the memory storing one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for executing the method for optimal MIMO Hybrid-beamforming under static and / or dynamic interference, as disclosed herein.

[0035] According to some embodiments, the device may be a cellular communication device, point to point MIMO wireless communication system and / or advanced WI-FI system.

[0036] According to some embodiments, there is provided a non-transitory computer-readable medium storing computer program instructions for executing the method for optimal MIMO Hybrid-beamforming under static and / or dynamic interference, as disclosed herein.

[0037] Certain embodiments of the present disclosure may include some, all, or none of the above advantages. One or more other technical advantages may be readily apparent to those skilled in the art from the figures, descriptions, and claims included herein. Moreover, while specific advantages have been enumerated above, various embodiments may include all, some, or none of the enumerated advantages.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. In case of conflict, the patent specification, including definitions, governs. As used herein, the indefinite articles “a” and “an” mean “at least one” or “one or more” unless the context clearly dictates otherwise.

[0039] BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Some embodiments are herein described, by way of example only, with reference to the accompanying drawings. With specific reference to the drawings in detail, it is stressed that the particulars shown are by way of example and for purposes of illustrative discussion of embodiments of the invention. In this regard, the description taken with the drawings makes apparent to those skilled in the art how embodiments of the invention may be practiced.

[0041] Attention is now directed to the drawings, where like reference numerals or characters indicate corresponding or like components. In the drawings:FIG. 1 shows a simplified schematic illustration of an exemplary multi HetNet wireless communication network, under multi-interferences;

[0042] FIG. 2 schematically shows a schematic illustration of generation of a training model for learning the non-linear function connecting AH and f(W,U), according to some embodiments;

[0043] FIG. 3 schematically shows a flow chart of an ML-based method for calculating an optimized combining matrix for optimal hybrid beamforming, according to some embodiments;

[0044] FIG. 4 shows a schematic block diagram of implementing the ML-based beamforming method with intelligent RAN system;

[0045] FIG. 5 shows line graphs representing the average channel capacity vs. signal interference ratio (SIR), as measured under various scenarios: Red line 510 represent the average capacity determined using the classical hybrid beam forming method, Blue line 520 represent the average capacity determined using the ML-based method, under the 1stexperiment of the 2ndarchitecture, and green line 530 represent the average capacity determined using the ML-based method, under the 2nd experiment of the 2ndarchitecture; and

[0046] FIGs. 6A-6B shows dot graphs of spatial power distribution in relation to calculating the channel’s eigenvalues as measured under the various scenarios: Red dots 610 represent the average capacity determined using the classical hybrid beam forming method (1stsimulation), Blue circles 620 represent the average capacity determined using the ML-based method, under the 1stexperiment of the 2ndarchitecture; green circles 630 represent the average capacity determined using the ML-based method, under the 2nd experiment of the 2ndarchitecture.

[0047] DETAILED DESCRIPTION

[0048] The following detailed description is of the best currently contemplated modes of carrying out the invention. The description is not to be taken in a limiting sense. Still, it is made merely the general principles of the invention, since the scope of the invention is best defined by the appended claims.

[0049] According to some embodiments, there are provided herein computerized systems and methods for enhanced MEMO Hybrid-beamforming under various interferences, wherein the computerized systems and methods make use of Machine Learning (ML) algorithms for enhancingthe channel matrix by high capacity and cancellation of interference, to thereby generate optimal MIMO-beamforming against static and dynamic interferences. In particular, the systems and methods provide a new technique of SIC-MIMO-beamforming without any estimation of CSI / CDI and physical feedback requirements between the receiver and the transmitter.

[0050] To facilitate understanding the context of the invention, the following general terms are herein briefly described:

[0051] Multiple Input Multiple Output (MIMO) technology is used primarily in wireless communications to enhance performance by utilizing multiple antennas at the transmitter and receiver ends. The main goal of MIMO is to improve data throughput, reliability, and network capacity without requiring additional bandwidth or increased transmission power. MIMO systems include multiple antennas used to transmit and / or receive signals, allowing spatial multiplexing, where different data streams are transmitted simultaneously over multiple spatial paths, effectively multiplying the data rate. Beamforming (directing the transmission or reception of signals in specific directions using antenna arrays) by MIMO systems is facilitated by adjusting the phase and amplitude of signals across antennas, thereby directing energy more efficiently to a receiver. MIMO can be used in various communication settings, including Wi-Fi, Cellular Networks, Satellite and Radio Communication, and the like.

[0052] Hybrid beamforming architecture combines digital beamforming and analog beamforming. Digital beamforming occurs in the baseband (digital domain), using digital signal processing. Analog beamforming occurs in the RF domain using analog phase shifters or switches and is used to adjust the phase of signals transmitted or received at each antenna to focus beams in specific directions. A typical hybrid beamforming system includes the baseband digital beamforming (handling multi-user precoding and postcoding), RF analog beamforming (controlling the phase and amplitude of the signals to form directional beams, using phase shifters, switches, or true timedelay circuits), antenna arrays (for transmitting and received the beamformed signals), and RF chains. In the transmit side hybrid beamforming a digital precoding is performed (processing baseband signals to form user-specific beams); analog RF beamforming is facilitated (shaping the signals into specific spatial patterns by adjusting phase shifts), the beamformed signals are transmitted using the antenna(s). In the receive-side hybrid beamforming, the antenna array(s) receive / captures signals, the RF domain adjusts the phase and amplitude to form received beams,and a digital postcoding is performed to combine the signals for decoding. Hybrid beamforming may be used in various wireless communication systems, particularly in millimeter-wave (mmWave) 5G networks and beyond, satellite communication, radar, and the like.

[0053] A base station is a central communication hub in wireless networks that connects user devices to the broader network infrastructure. It acts as a transmitter and receiver, facilitating communication between user equipment (UE) and the core network. The base station transmits data to user devices via downlink (base station to user) and receives data from user devices via uplink (user to base station). The base station usually includes antenna arrays, Radio Frequency (RF) components (such as amplifiers, filters, and converters to handle RF signals), digital processing unit (performing signal processing, including modulation, encoding, and beamforming), power amplifier (configured to amplify signals before transmission to ensure sufficient coverage) and / or control and interface Unit.

[0054] A channel matrix is a mathematical representation of the communication channel in the MIMO system. It generally describes how signals are transmitted between multiple transmitting and receiving antennas, capturing the effects of the wireless environment. The channel matrix may be denoted as H; the channel matrix is typically an Nr times Nt matrix, wherein Nt is the number of transmit antennas, and Nr is the number of receive antennas. Each element hq represents the channel coefficient between the jthtransmit antenna and the ithreceive antenna. This coefficient accounts for Path loss, Shadowing, Multipath effects (reflection, scattering, and diffraction), and Fading characteristics. In a MIMO system, the received signal vector, y, can be expressed as y=Hx+n, wherein x is the transmitted signal vector, H is the channel matrix, and n is the noise vector. Channel matrices include deterministic, random, and sparse channel matrices. The combining matrix (U) is used at the receiver to process the incoming signals from multiple antennas. It transforms the received signal vector into a form suitable for detection or further processing. The initial linear combining matrix refers to the starting or basic form of the combining matrix U. At the initial stage, this matrix is designed to linearly transform the received signal vector y into a representation that optimizes the detection of transmitted signals x. In some embodiments, interferences in communication may be designated by W. Such interferences include static interferences. In some instances, Z may represent a white noise and / or dynamic interference, such as additive white Gaussian noise (AWGN).According to some embodiments, the disclosed systems and methods may advantageously be used to overcome one or more of the following communication related issues, including: multipath effects; cancelling effects of static interference(s) (such as, scatters, reflectors, static users, or jamming devices); cancelling effects of dynamic interference (such as, dynamic positions of devices or jamming device), cancelling effects of mutual interference from different HetNet cells; dealing with passive and active jammers randomly deployed in space; dynamic multi-user MIMO problems (uplink and downlink); scenarios where CDI, singular value decomposition (S.V.D) process and / or advanced SIC (such as, NOMA, MMSE-IRC, etc.) are ineffective; optimal balance between cancelling interference and increasing the channel’s capacity, multy-layer beamforming spatial-multiplexing decoding problems, direction of arrival (DOA) 3D array estimation response and multi-carrier channels issues, and the like, or any combinations thereof.

[0055] Reference is now made to Fig. 1, showing a simplified schematic illustration of an exemplary multi-HetNet wireless communication network under multi-interferences. Generally, a heterogeneous network (HetNet) is a communication network that includes various types of cells or base stations (BS) with varying sizes, power levels, and coverage areas. As shown in Fig. 1, Exemplary HetNet 100 includes a plurality of base stations (such as base stations 120A-F, each having its wireless coverage area (marked by a circle or ellipsoid)) and end devices (such as end device 130A-G). The base stations and / or end devices are located in various areal locations (110A-C), wherein each location has a variety of interferences. Such interferences may include static interferences (generally referred to as W) or dynamic interferences (generally referred to as Z), and may be generated by various causes, including, other base station, end devices, jammers, environmental conditions (weather, geographical conditions, and the like). Each of the base stations has / can generate its own channel matrix (H) and a combining matrix (U) to facilitate communication. For example, base station 1 has a channel matrix Hi, can generate combining matrix Ui, and is under interference Wi. In order to improve communication between various components of the HetNet (or other multi-users communication networks), reducing the interferences should be facilitated, in order to generate an optimal combining matrix between transmitters and receivers.

[0056] According to some embodiments, there are provided herein methods and systems that can enhance and optimize communication under various interferences, by affecting beamforming, in order to generate an optimal combining matrix. In certain embodiments, the systems and methodsdisclosed herein utilize machine learning tools and other algorithms to enable the real-time online calculation and / or determination of an optimal combining matrix. This approach aims to enhance communication efficiency.

[0057] According to some embodiments, advantageously, as exemplified herein, the systems and methods disclosed herein facilitate self-computational feedback on the receiver side only, thereby increasing efficiency, accuracy, and security.

[0058] In some embodiments, the systems and methods demonstrate significant improvements in MIMO channel performance, specifically regarding capacity and optimal hybrid beamforming. According to some embodiments, as exemplified herein, utilizing ML tools advantageously enables solving multi-objective optimization, facilitating the solution of various problems as a generalization of one overall problem. This is in contrast to the classical approaches with a matrix design approach in a single combination, designed according to one particular issue.

[0059] Reference is now made to Fig.2, which shows a schematic illustration of the generation of a training model for learning a non-linear function f(W,U) for modeling the interference term AH, according to some embodiments. As shown in Fig. 2, for the generation of the training model, a transmitter-receiver system 200 is used. To this aim, transmitter side 210 generates a plurality of signals under various interference scenarios, wherein for each scenario, a non-linear function correlating an interference (W) and a combining matrix (U) is generated based on the received signal on receiver 220. As shown in Fig. 2, transmitter 210 transmits a radio band (RB) signal S, which is received as RB signal Y at receiver 220. HTR is the wireless channel matrix from the transmitter to the receiver, and HJR is the wireless channel matrix from the interference to the receiver. W is the static interference, U is the combining matrix and Z is the AWGN or a dynamic interference. Based on the differences in the transmitted and the received signals, under various interferences / conditions, the AH non-linear function, correlating W and U can be determined. The determination may be performed off-line, as part of a training and learning module. Thereafter, based on the trained function, the receiver can determine, on-line, in real time (for example, using a processing module, configured to execute an algorithm to determine an optimal U, based on the non-linear function AH, as detailed below herein), an optimal combining matrix, suitable for the specific real time condition. According to some exemplary embodiments, the model may be calculated according to the equation: Y= (HTR+ AH)S + Z = (HTR + f(W, U))S+ Z.Reference is now made to Fig. 3, a flow chart of an ML-based method for calculating an optimized combining matrix for hybrid beamforming, according to some embodiments. As shown in Fig. 3, method 300 includes an offline part (steps 310-330), in which the model is trained to learn / determine the non-linear function f(W,U), as detailed herein. In addition, the method includes an online part (steps 340-380), in which, based on the learned function, a final, optimal combining matrix is calculated at the receiver. In the off-line part, as shown in step 310, data from various interfering areas of different interference is collected, for example, using model-based transmitter-receiver classical hybrid-beamforming. The training set may include 100-100000 or more multi-path MIMO channels, including various passive and active jammers randomly deployed in the space. The training set may consider non-separable linearly complex patterns of wireless MIMO channels combined with various interference and destructive effects. The collected interference data is then clustered. The clustering may be performed by various suitable means, including, for example, the K-means classification algorithm. The classification process results for each cluster with a central point and radius for the smallest matrix ball containing all the data-set points clustered to this cluster. As shown in step 320, the clusters are then represented by an image of interference matrix W, and the positioning of the receiver combining matrix U, with a characteristic radius for each cluster, where their joint influence is recorded as the related signal at the receiver. In some embodiments, the joint influence may be modeled through an artificial neural network, for example, radial basis functions (RBF) model. Next, at step 330, based on the represented imaged clusters, the non-linear function f(W,U), connecting W and U to AH, the error function of the interpolation process is learned and determined. Thus, the results of the off-line part (training), can be used for determination of an optimal final combining matrix calculation, at the receiver side, by executing the steps of the on-line part of method 300. As detailed herein, the on-line part is executed at the receiver, enabling the receiver to approximate an initial linear model of the combining matrix, the channel-MIMO wireless-matrix, and the interference, updating the algorithm, and based on the learning-training process, identify the correct channel matrix and approximate the optimal combining matrix for the unknown on-line, real-time situation, in which it should be positioned. Thus, at step 340, the method includes obtaining / getting the initial receiving matrix signal (Resource-Block Y) and the initial combining beamforming matrix (U). At step 350, initial static and dynamic interference (W) and the weight matrix (B) of the combining matrix is calculated. Next, at step 360, an initial optimal combiningmatrix is calculated, based on steps 340-350. At step 370, over iterations of the non-linear function (AH=f(W,U)) are performed, in order to get an exact and final interference finger-print (W) and the weight matrix (B). Next, on step 370, based on the determined final W and B, an optimal and final combining RF and Baseband matrix is calculated.

[0060] According to some embodiments, the training set may be created using model-based transmitter-receiver classical hybrid-beamforming and model-based interferences to train the model. These channels also depend on the design of pre-coding / combination matrices of the Radio-frequency (RF) and Base-Band (BB) in hybrid beamforming techniques. Hence, this dependence may be considered when forming the training set.

[0061] In some embodiments, the training set may include collecting data (CSI, CDI, interferences) from the frequency domain between the RU and DU from different interfering areas and saving the database in the PXI component.

[0062] In some embodiments, the training set may consider non-linearly-separable complex patterns of wireless MIMO channels combined with various interference and destructive effects. The wireless-MIMO channel matrix may depend nonlinearly on the interferences, on the fast dynamic of the destructive effects and receiver motion, and on the combining matrix of the receiver. Thus, the challenge in such a situation is to prepare the receiver ML hybrid-beamforming to operate in real-time online under a real physical wireless environment for identification of the correct wireless MIMO channel matrix and approximate the optimal combining matrix at the receiver through self-computational optimal feedback based on the learning, training, and testing offline processes.

[0063] According to some embodiments, the training step of the model which includes classifying the set of non-linearly-separable patterns (the channel-MIMO-wireless matrices and the combining matrices sets) into separable ones, can be facilitated, for example, with K-means algorithm, where minimum K-clusters are approximated for all the records of the training set. The classification process results for each cluster with a central point and radius for the smallest matrix ball containing all the data-set points clustered to this cluster. Each point in the cluster contains an interference matrix and a combining matrix, where their joint influence is recorded as the related signal at the receiver. The joint influence may be modeled through an artificial neural network, for example, the Radial Basis Functions (RBF) model, such as RBFNN, where the radial basisfunctions are Gaussian functions, and each cluster is modeled through a single Gaussian. The weights may be computed to minimize the Mean- Square-Error (MSE).

[0064] According to some embodiments, the off-line and on-line parts of method 300 can be performed on the same or separate computational modules. In some embodiments, the online part may be performed separately (in time and / or space and / or computational resources) from the offline training part. In some embodiments, the training may be performed / executed offline and stored on a local or remote database / memory module. The offline part may be continuously generated and periodically updated in some embodiments. In some embodiments, the offline part may be updated in real-time.

[0065] According to some embodiments, the model is trained in a supervised manner.

[0066] In some embodiments, the learning algorithm may utilize Radial Basis Functions Neural Networks (RBFNN) with the K-means algorithm and Recursive Least Squares (RLS) algorithm to gain fast computations and fast updates.

[0067] According to some embodiments, classification algorithms for clustering may include, for example, K-means, Gaussian Mixture Models (GMM), DBSCAN (Density-Based Spatial Clustering of Applications with Noise), Mean Shift Clustering, Fuzzy c-Means, and the like, or any combinations thereof. Each possibility is a separate embodiment.

[0068] According to some embodiments, classification algorithms may include, for example, Radial Basis Function Neural Networks, Support Vector Machines (SVM) with RBF Kernel, Gaussian Process Regression (GPR), Multilayer Perceptron (MLP), k-Nearest Neighbors (k-NN), Decision Trees and Random Forests, Extreme Learning Machine (ELM), Adaptive Boosting (AdaBoost), Convolutional Neural Networks (CNNs), Kernel Ridge Regression (KRR), Deep Gaussian Processes (DGPs), and the like, or any combinations thereof. Each possibility is a separate embodiment.

[0069] According to some embodiments, once the model is trained, it may be applied to the online process. To this aim, the hybrid-beamforming receiver-based ML may be placed under an unknown and random challenger wireless-interference environment(s). A classical hybridbeamforming transmitter transmits heavy information through random direction beams to the hybrid-beamforming receiver-based ML. The receiver approximates, using the disclosedalgorithms, the initial linear approximation of the combining matrix, the channel-MIMO wirelessmatrix, and the interference, updating the algorithm, and based on the learning -training process, can identify the correctness channel matrix and approximate the optimal combining matrix for this unknown situation.

[0070] According to some embodiments, the ML-based hybrid beamforming may be implemented under a limited learning set and minimum clusters. According to some embodiments, as exemplified below, the volume of information capacity obtained using the RBFNN is significantly higher than that obtained using the conventional way of creating hybrid beamforming. According to some embodiments, the efficiency of replacing the conventional decoding mechanism of the classical hybrid beamforming at the receiver with ML tools is drastically enhanced.

[0071] According to some embodiments, the methods disclosed herein may be integrated, in the form of algorithms in components of a base station to thereby produce techniques of new Hybrid-Beamforming based ML in mmWave (for example, in the range of 6-100GHz), as well as new digital SIC based on ML in MEMO scheme below 6GHz.

[0072] According to some embodiments, the methods and algorithms disclosed herein may be integrated in various settings. For example, the algorithms may be integrated in RAN (Radio Access Network) systems, to improve efficiency, performance, and adaptability. Intelligent RAN may be used, for example, in 5G and 6G networks.

[0073] Reference is now made to Fig- 4, which illustrates a schematic block diagram of implementing ML-based beamforming with an intelligent RAN system, according to some embodiments. As shown in Fig. 4, intelligent Radio Access Network (RAN) 400 includes a container ML training module 410, configured to train / store the trained model (off-line process 412, including steps 1-5, which are equivalent to steps 310-330 of method 300 in Fig. 3). RAN 400 includes an Intelligent Controller (RIC), which is configured to provide a framework for executing the AI / ML-based methods. The RIC of RAN 400 interacts with e / g node B component 420. The Evolved Node B (eNodeB) and gNodeB are components in the architecture of 4G LTE and 5G networks, respectively. They serve as the connection points between the user equipment (UE) and the core network / base station. The e / g node B component includes a centralized unit (CU 430) configured to handle near-real-time tasks and higher-layer protocols, including, for example, performing Radio Resource Control (RRC), mobility management, and QoSenforcement and managing connection to the core network and other base stations. The node further includes a real-time distributed unit (DU 440), which is configured to handle real-time processing tasks, such as Medium Access Control (MAC), and Physical (PHY) layer operations like encoding, decoding, and scheduling. The node further includes a Radio Unit (RU 450) configured to deal with the radio frequency (RF) front end, including Signal transmission and reception, Conversion between analog and digital signals, and Beamforming using Massive MIMO. The node further includes Front Distribution (FD unit 460), which is the network connection between the RU and DU and is configured to transport radio signals in digitized form and implement synchronization for coordinated operations. As shown in Fig. 4, the e / g / node component is capable of executing the algorithm to determine an optimal combining matrix (online process 416, including steps 6-8, equivalent to steps 340-380 of method 300 in Fig. 3). The e / gNodeB components (CU, DU, and RU) interact with the RIC (for example, Near-real-time (RT) RIC communicates with the DU to enable real-time optimizations, and non-RT RIC communicates with the CU for long-term policy enforcement and AI / ML integration. Thus, the architecture at the e / g-node B can be modified to establish a cognitive ability, based on the ML, using the disclosed algorithms and methods, of implementation of base stations, in order to enhance and optimize wireless communication.

[0074] According to some embodiments, the most destructive communication scenarios for advanced MIMO wireless communication systems are scenarios in which the image of the interferers changes rapidly in relation to the S.V.D decomposition process and the beamforming process. Another complex procedure is where both the end users and the interferers change their spatial location dynamically, or when the medium has a high amount of scatterers so that even if the receiver-transmitter manages to shape beams, they scatter, a process that causes significant degradation in system performance. Traditional SIC and traditional Hybrid-beamforming methods have difficulty dealing with these complex scenarios. To test and simulate the efficiency of the algorithms and the architecture of the ML Hybrid-beamforming, simulators may be used. For example, a control simulator may include a transceiver based on a classic Hybrid-beamforming architecture, including, for example, a modulation based on QPSK modulation, a baseband block including samples, and a digital BB pre-coding matrix, FBBTR. After these blocks, four blocks are created based on four RF chains. For each block, eight transmission antennas (the total transmission antennas are Nt = 32) may be defined, where a phased array is connected to eachtransmission antenna and controlled by an analog-RF-pr ecoding matrix, FRFTR. On the receiver side, an array of Nr = 32 and thirty -two phase shifters may connect thereto, controlled with an analog-RF-combining matrix, WRFTR. The ability to estimate the DO A of the beams may be based on the MUSIC algorithm. The demodulate RF chains and the BB decoding process may be symmetric to the transmitter, with BB combining matrix, WBBTR. For example, a test simulator may include the architecture as the control simulator. Still, on the receiver side, the front end of the receiver is replaced with an algorithmic array based on the ML method (including RBFNN and K-means with the output of the algorithms, including exporting clusters in which optimal RF-combining matrices are calculated with the weight matrix for each MEMO wireless channel matrix scenario). Running the ML-based architecture includes two main stages - an offline calculation process - the learning process, and an online calculation process, as detailed above herein. Each step includes several main calculation steps. The first step in the offline process is based on collecting data from different interfering areas and calculating K-means to obtain clusters. The clusters are each represented by an image of interference and the receiver’s positioning, with a characteristic radius for each cluster. The last step in the offline process is based on the learning process, with learning the nonlinear function, NH =f (W,U), representing the interferences. The online process includes several calculation steps. The first step is getting the initial received signal, Y, and the initial RF-combining matrix, the second step is to calculate initial interferences, W, and the weight matrix B. The third step is to initiate the optimal RF-combining matrix calculation, and over iterations, AH = f (W,U), to get the final and the exact weight matrix B. The final step is to get the optimal RF-combing matrix. The simulators may be used to determine various mapped parameters at the receiver, for example, the average channel capacity and the spatial power gain divided according to the calculation of the eigenvalues of the channel through a classical S.V.D. decomposition. Based on the measurement of these parameters under various conditions, the efficiency of the hybrid beamforming may be evaluated. As detailed herein below, the ML-based hybrid beamforming can achieve up to lObits / s / Hz under SIR = 5dB (under random interference) or up to 30bits / s / Hz under SIR = 5dB (under trained interferences), compared to the classical Hybrid Beamforming. Moreover, in various ranges, such as between -20dB < SIR < -15dB, the ML-Hybrid-Beamforming can achieve 15 bits / s / Hz and 10 bits / s / Hz (under-trained or random interferences, respectively). Moreover, when using the ML-based hybrid beamforming method, the power concentration is much more efficient than that of the classical hybrid beamformingmethod. For example, based on the ML-hybrid beamforming method, the power is concentrated in certain eigenvalues compared to the classical hybrid beamforming technique, in which the power distribution in some eigenvalues is wasteful.

[0075] Some embodiments, there is provided a method for hybrid beamforming that involves calculating an optimal combining matrix at a receiver using machine learning tools based on a trained algorithm. In some embodiments, the method is an ML-based approach for optimal MIMO hybrid beamforming, designed to address interference and jamming attacks (both static and dynamic) in order to improve the quality of the received signal. In some embodiments, theThe method includes both an offline training phase and an online execution phase.

[0076] According to some embodiments, the offline part may include the steps of:

[0077] a) Creating a database (object storage) from the RF-IF and Baseband Domain unit. The database can also be generated to make the information as rich as possible. For fixed transmitted symbol matrices Si, i=l,...,N and fixed combining matrices Uj, j=l,...,M, using the received signal Yij, the disrupted channel matrix Hij, is computed, the dynamic interference Zij, the static interference Wij and the weight matrix Bij for the influence of the combining matrix Uj on Yij. Data patterns may be created from various interfering areas (including interference and destructive effects), the data patterns may include channel MIMO-wireless interference matrices (AHi, 1=1,..., L), static interference matrices (Wi), and control combining matrices (Ui);

[0078] b) classifying the collected data patterns to obtain one or more clusters, for example, using (K-means algorithm, based on Wij and Uj), wherein each point in a cluster includes a static interference matrix (Wij) and a control combining matrix (Uj) data set points, having a similar joint influence on the received signal (Yij), wherein the joint influence effect is determined based on a weight matrix B, , ;

[0079] c) determining, using an ML algorithm, a non-linear function model for the AH=f(W,U), correlating the static interference matrix (W) and the combining matrix (U). According to some embodiments, the online part may include the following steps: a) receiving, by a receiver unit, an initial received wireless signal (Y), transmitted from a transmitter (transmitting signal (S)) and an estimated combining matrix (estimated Uest);b) calculating / estimating an initial interference matrix Wmi and a weight matrix Bini, based on the initial Y and the estimated Uest;

[0080] c) calculating, using an ML algorithm, based on the calculated initial Wmi and Bini, an initial optimal Uiopt;

[0081] d) calculating, using an ML algorithm, based on one or more iterations of the determined non-linear function, a final W (Wfin) and a final B (Bfin), to thereby determine a final optimal Ufopt, where in each step, the combining matrix is corrected according to a closed-form formula for the optimal U;

[0082] e) adjusting / adapting the receiver beamforming based on the Ufopt, thereby optimizing the received signal.

[0083] According to some embodiments, the hybrid beamforming technology that utilizes machine learning techniques enables the design of multiple beamformers. It directs the main beamforming gain towards the user locations, while simultaneously employing null-steering to target interference sources. This functionality operates effectively even in challenging propagation environments, such as selective-frequency transfer characteristics and non-stationary channels, as well as in dynamic channel conditions.

[0084] According to some embodiments, the method can be applied across various protocols and international standards in authorized networks, including LTE, 5G-NR, mobile 6th generation 6G, Wi-Fi 6-9, Bluetooth, satellite communication, GPS, GLONASS, and Galileo. Additionally, this method applies to standalone unlicensed systems, point-to-point MEMO systems, and unlicensed networks. Each possibility is a separate embodiment.

[0085] According to some embodiments, the method can be applied in frequency ranges below 6 GHz and in the millimeter wave range of about 6-100 GHz. Each possibility is a separate embodiment.

[0086] According to some embodiments, the optimization / enhancement of the received wireless signal may be facilitated only on the receiver side.

[0087] According to some embodiments, the optimization / enhancement of the received wireless signal does not require feedback, any other closed-loop-MIMO, or any operation of Physical Upload Share Channels (PUSCH) between the transmitter and the receiver.According to some embodiments, the classification may utilize an RBFNN and K-means algorithm. In some embodiments, the non-linear model of AH can be any non-linear fast ML model or any Al model with few hidden layers. In some embodiments, the classification method can be any method for classifying numerical data according to similarity related to some distance function.

[0088] According to some embodiments, the receiver unit may be a base station, an e / g-nodeB (for example, based on ORAN 7.2), a relay, a small cell, or any point-to-point transceiver. Each possibility is a separate embodiment. According to some embodiments, the receiver unit may be an end unit (such as a smartphone, Communication device, etc.).

[0089] According to some embodiments, the methods and systems disclosed herein may be implemented under various settings, including, for example, the industrial sector (for example, in 5G-NR networks, for example, ORAN 7.2).

[0090] According to some embodiments, the computerized models disclosed herein may utilize Machine learning (ML) and Artificial intelligence (Al) tools, including any type of suitable algorithms, such as, for example, but not limited to: transformers, artificial neural network(s) (ANN), such as convolutional neural network (CNN), recurrent neural network (RNN), long-short term memory (LSTM), auto-encoder (AE), generative adversarial network (GAN), Reinforcement-Learning (RL), support vector machine (SVM), decision tree (DT), random forest (RF), and the like. Both “supervised” and “unsupervised” methods may be implemented.

[0091] According to some embodiments, there is provided a non-transitory memory device, wherein modules of instruction code are stored, and at least one processor associated with the memory device, and configured to execute the modules of instruction code, whereupon execution of said modules of instruction code, the at least one processor is configured to execute one or more of the methods disclosed herein, for calculating an optimal combining matrix.

[0092] According to some embodiments, there is provided a system for calculating an optimal combining matrix, the system includes one or more processors and, optionally, RAM and / or nonvolatile memory components associated with the one or more processors, the processors are configured to execute the method of calculating or generating an optimal combining matrix.According to some embodiments, there is provided a computer-readable storage medium having stored therein machine learning software, executable by one or more processors for calculating an optimal combining matrix.

[0093] According to some embodiments, there is provided a computer-implemented method for optimizing MIMO hybrid beamforming under static and / or dynamic interference, including: (a) receiving, by a receiver unit, an initial received wireless signal Y, transmitted from a transmitter sending a signal S, and receiving an estimated combining matrix U_est;

[0094] (b) calculating, based on Y and U est, an initial interference matrix W inl and a weight matrix B ini;

[0095] (c) determining, using a machine learning algorithm trained to model a non-linear mapping function AH correlating an interference matrix W and a combining matrix U, an initial optimal combining matrix U iopt, based on W inl and B ini;

[0096] (d) iteratively calculating, using said machine learning algorithm, updated interference matrices and weight matrices to obtain a final interference matrix W fin, a final weight matrix B fin, and a final optimal combining matrix U fopt, wherein in each iteration the combining matrix is corrected using a predefined closed-form formula for optimal combining; and

[0097] (e) adjusting beamforming of the receiver unit based on U fopt, thereby optimizing reception of the wireless signal.

[0098] According to some embodiments, the machine learning algorithm is executed by the receiver unit.

[0099] According to some embodiments, the method is performed in real-time with a processing latency below a predefined threshold.

[0100] According to some embodiments, the wireless signal may include one or more of: LTE, 5G-NR, mobile 6th generation, Wi-Fi 6, Wi-Fi 7, Wi-Fi 8, Wi-Fi 9, Bluetooth, satellite communication signal, GPS, GLONASS, and Galileo signal.

[0101] According to some embodiments, the wireless signal frequency is below about 6 GHz, or in the range of about 6 GHz to 100 GHz.

[0102] According to some embodiments, the optimization of the received wireless signal is performed exclusively at the receiver unit.According to some embodiments, the method is performed without requiring uplink feedback, closed-loop MIMO control, or Physical Uplink Shared Channel (PUSCH) signaling from the receiver to the transmitter.

[0103] According to some embodiments, the receiver unit may be selected from a base station, an e / g-nodeB component, a relay, a small cell, or a point-to-point transceiver.

[0104] According to some embodiments, the receiver unit is an end-user device.

[0105] According to some embodiments, the machine learning algorithm is trained offline using a database including a plurality of data patterns, each comprising a static interference matrix W_i,j, a control combining matrix UJ, and a received signal Y_i,j, with an associated weight matrix B_i,j representing their joint influence on signal quality, and wherein said training determines a nonlinear function AH correlating W and U.

[0106] According to some embodiments, the data patterns may be derived from different interference environments and comprise channel MIMO interference matrices AH I, static interference matrices W_l, and control combining matrices U_l.

[0107] According to some embodiments, the interference may include one or more of: interference signals from general users, interference signals from smart jamming devices, random scattering objects in wireless channels, substantial multipath effects, active jammers, and passive jammers.

[0108] According to some embodiments, there is provided a system for optimizing MIMO hybrid beamforming under interference, including a processor configured to execute the method as disclosed herein.

[0109] According to some embodiments, the system may further include a receiver unit functionally associated with the processor.

[0110] According to some embodiments, there is provided an electronic device including one or more processors; and memory coupled to the one or more processors, the memory storing one or more programs including instructions which, when executed by the one or more processors, cause the device to perform the method disclosed herein.

[0111] According to some embodiments, the device is a cellular communication device, a point-to-point MIMO wireless communication system, or an advanced Wi-Fi system.According to some embodiments, there is provided a non-transitory computer-readable medium storing computer program instructions which, when executed by one or more processors, cause performance of the method as disclosed herein.

[0112] In the description and claims of the application, the words “include” and “have” and forms thereof, are not limited to members in a list with which the words may be associated.

[0113] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. In case of conflict, the patent specification, including definitions, governs. As used herein, the indefinite articles “a” and “an” mean “at least one” or “one or more” unless the context clearly dictates otherwise.

[0114] It is appreciated that certain features of the disclosure, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the disclosure, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable subcombination or as suitable in any other described embodiment of the disclosure. No feature described in the context of an embodiment is to be considered an essential feature of that embodiment unless explicitly specified as such.

[0115] Although stages of methods according to some embodiments may be described in a specific sequence, methods of the disclosure may include some or all of the described stages carried out in a different order. A method of the disclosure may include a few of the stages described or all of the stages described. No particular stage in a disclosed method is to be considered an essential stage of that method, unless explicitly specified as such.

[0116] Although the disclosure is described in conjunction with specific embodiments thereof, it is evident that numerous alternatives, modifications, and variations that are apparent to those skilled in the art may exist. Accordingly, the disclosure embraces all such alternatives, modifications, and variations that fall within the scope of the appended claims. It is to be understood that the disclosure is not necessarily limited in its application to the construction details and the arrangement of the components and / or methods set forth herein. Other embodiments may be practiced, and an embodiment may be carried out in various ways.The present invention may be a system, a method, and / or a computer program product. The computer program product may include a computer-readable storage medium (or media) having computer-readable program instructions thereon for causing a processor to carry out aspects of the present invention.

[0117] A computer program (also referred to as a program, software, software application, script, or code) can be written in any form of programming language, including compiled or interpreted languages and declarative or procedural languages. It can be deployed in any form, including as a standalone program or as a module, component, subroutine, object, or other unit suitable for use in a computing environment. A computer program may, but need not, correspond to a system file. A computer program can be stored in a portion of a file that holds other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (for example, files that store one or more modules, sub-programs, or portions of code). A computer program can be deployed to be executed on one computer or multiple computers located at one site or distributed across multiple sites and interconnected by a communication network.

[0118] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer-readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium within the respective computing / processing device.

[0119] Computer readable program instructions for carrying out operations of the present invention may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, for example, JavaScript, Smalltalk, C, C++, TypeScript, Python and R.

[0120] The computer-readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer,and partly on a remote computer or entirely on the remote computer or server (such as a cloudbased). In the latter scenario, the remote computer (or cloud) may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider) including wired or wireless connection (such as, for example, Wi-Fi, BT, mobile, and the like). In some embodiments, electronic circuitry, including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA), may execute the computer-readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present invention. Moreover, a computer can be embedded in another device, for example, a mobile phone, a tablet, a personal digital assistant (PDA), or a portable storage device (for example, a USB flash drive). Devices suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, including semiconductor memory devices, for example, EPROM, EEPROM, random access memories (RAMs), including SRAM, DRAM, embedded DRAM (eDRAM) and Hybrid Memory Cube (HMC), and flash memory devices; magnetic discs, for example, internal hard discs or removable discs; magneto-optical discs; read-only memories (ROMs), including CD-ROM and DVD-ROM discs; solid state drives (SSDs); and cloud-based storage. The processor and the memory can be supplemented by or incorporated into specialpurpose logic circuitry.

[0121] Aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams and combinations of blocks in the flowchart illustrations and / or block diagrams can be implemented by computer-readable program instructions.

[0122] These computer-readable program instructions may be provided to a processor of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer-readable program instructions may also be stored in a computer-readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer-readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0123] The computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0124] The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.

[0125] The processes and logic flows described herein may be performed in whole or in part in a cloud computing environment. For example, some or all of a given disclosed process may be executed by a secure cloud-based system comprised of co-located and / or geographically distributed server systems. The term “cloud computing” is generally used to describe a computing model which enables on-demand access to a shared pool of computing resources, such as computernetworks, servers, software applications, and services, and which allows for rapid provisioning and release of resources with minimal management effort or service provider interaction.

[0126] The descriptions of the various embodiments of the present invention have been presented for purposes of illustration, but are not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein was chosen to explain best the principles of the embodiments, the practical application or technical improvement over technologies found in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

[0127] While certain embodiments of the invention have been illustrated and described, it will be clear that the invention is not limited to the embodiments described herein. Numerous modifications, changes, variations, substitutions and equivalents will be apparent to those skilled in the art without departing from the spirit and scope of the present invention as described by the claims which follow.

[0128] EXAMPLES

[0129] As detailed herein, the most destructive communication scenarios for advanced MIMO wireless communication systems are scenarios in which the identity and visibility of the interferers change rapidly in relation to the S.V.D decomposition process and the beamforming process.

[0130] Another complex scenario is where both the end users and the interferers change their spatial location dynamically, or when the medium has a high amount of scatterers so that even if the receiver-transmitter manages to shape beams, they scatter, a process that causes significant degradation in system performance. Traditional SIC and traditional Hybrid-beamforming methods have difficulty dealing with these complex scenarios.

[0131]

[0132] The following simulators were developed to test and simulate the efficiency of the disclosed algorithms for ML Hybrid-beamforming, including two different architectures.In the first architecture, a transceiver was created based on classic Hybrid-beamforming architecture, according to the following: In the transmitter, a modulation was designed based on QPSK modulation, a baseband block including samples, and digital BB pre-coding matrix, FBBTR. After these blocks, four blocks were created based on four RF chains. For each block, eight transmission antennas were defined (the total transmission antennas are Nt = 32), where a phased array is connected to each transmission antenna and controlled by an analog-RF-precoding matrix, FRFTR. On the receiver side, an array of Nr = 32 and thirty-two phase shifters are connected and controlled with the analog -RF-combining matrix, WRFTR. The ability to estimate the DO A of the beams is based on the MUSIC algorithm. The demodulate RF chains and the BB decoding process are symmetric to the transmitter, with BB combining matrix, WBBTR. In this simulation, a dataset of one hundred and twenty-five MIMO wireless channel matrices were created based on quasistatic-fading, static effects such as scatters, and dynamic interferences. Experiments were executed with draws of randomly transmitted spatial angles, DOD, and randomly generate a wireless channel matrix from the set of measurements compiled.

[0133] On the receiver side, two main parameters were measured, the average channel capacity and the spatial power gain divided according to the calculation of the eigenvalues of the channel via a classical S.V.D. decomposition.

[0134] In the second architecture, the block diagram of the transmitter is precisely the same as the block diagram of the first architecture. On the receiver side, the front end of the receiver is replaced with an algorithmic array based on RBFNN and K-means with the output of the algorithms, exporting clusters in which optimal RF-combining matrices are calculated with the weight matrix for each MIMO wireless channel matrix scenario. The number of the receiver antenna array and the number of phase shifters are the same as in the first architecture. The continuation of the receiver scheme of the decoding network in the receiver is the same as the first architecture.

[0135] The ML-based architecture process includes two main stages - an offline calculation process, which consists of the learning process, and an online calculation process. Each step includes several main calculation steps. The offline process's first step is collecting data from different interfering areas and calculating K-means to obtain clusters. The clusters are eachrepresented by a fingerprint of interference and the receiver’s positioning, with a characteristic radius for each cluster.

[0136] The last step in the offline process is based on the learning process, with learning the nonlinear function, H=f (W,U), representing the interference's influence on the received signal.

[0137] The online process includes several calculation steps. The first step is getting the initial of the received signal, Y, and the RF-combining matrix, the second step is to calculate initial interferences, W, and the weight matrix B. The third step is to initiate the optimal RF-combining matrix calculation, and over iterations, AH = f (W,U), to get the final and the exact weight matrix B. The final step is to get the optimal RF-combing matrix.

[0138] In the second architecture, two series (simulation) of experiments were conducted: In the first series (simulation), a wireless channel matrix is generated from the data set and the average channel capacity and the spatial power distribution in relation to calculating the channel’s eigenvalues are measured. In the second series (simulation) of experiments, a wireless channel matrix based is produced on fingerprints of interferers that are completely different from the existing data set. Under this scenario, the system performance is measured by determining the average channel capacity and spatial power distribution in relation to the classical Hybridbeamforming channel capacity and spatial power distribution.

[0139] The results are presented in Figs. 5 and 6A-B, showing the results of the average channel capacity and the spatial power distribution in relation to calculating the channel’s spatial power distribution, respectively. The line graphs shown in Fig. 5 represent the average channel capacity vs. signal interference ratio (SIR), as measured under the various scenarios: Red line 510 represent the average capacity determined using the classical hybrid beamforming method (1stsimulation (architecture), Blue line 520 represent the average capacity determined using the ML-based method, under the 1stexperiment of the 2ndarchitecture, and green line 530 represent the average capacity determined using the ML-based method, under the 2nd experiment of the 2ndarchitecture. Likewise, the dot graphs shown in Fig. 6A represent the spatial power distribution in relation to calculating the channel’s eigenvalues as measured under the various scenarios: Red dots 610 represent the average capacity determined using the classical hybrid beamforming method (1stsimulation), Blue circles 620 represent the average capacity determined using the ML-basedmethod, under the 1stexperiment of the 2ndarchitecture. The dot graphs shown in Fig.6B represent the spatial power distribution in relation to calculating the channel’s eigenvalues: Red dots 610 represent the average capacity determined using the classical hybrid beam forming method (1stsimulation), and green circles 630 represent the average capacity determined using the ML-based method, under the 2nd experiment of the 2ndarchitecture.

[0140] As can be seen from the results, the ML-Hybrid-Beamforming achieves up to lObits / s / Hz under SIR = 5dB in the second scenario and up to 30bits / s / Hz under SIR = 5dB in the first scenario, compared to the classical Hybrid Beamforming. Moreover, as can be deduced from Fig. 5, in the ranges between -20dB < SIR < -15dB, the ML-Hybrid-Beamforming achieves 15bits / s / Hz and 1 Obits / s / Hz, in the first scenario and in the second scenario, respectively. Further, as the SIR values increase, the green and blue curves rise sharply compared to the red curve.

[0141] Moreover, as can be seen in Figs.6A-B, it is possible to distinguish the spatial distribution of powers according to the eigenvalues of the channel matrices. From these illustrations, it is possible to notice the trend that the ML algorithm, based on the training and learning processes, determines to concentrate the power in certain eigenvalues compared to the classical technique, in which the distribution of the power in a certain eigenvalue turns out to be wasteful for example in the eigenvalue number 15 in Fig. 5.

Claims

CLAIMSWHAT IS CLAIMED IS:

1. A computer-based method for optimal MIMO Hybrid-beamforming under static and / or dynamic interference to optimize a received wireless signal transmitted from a transmitter, the method comprising:a. receiving, by a receiver unit, an initial received wireless signal (Y), transmitted from a transmitter, transmitting a signal (S), and an estimated combining matrix (estimated Uest);b. calculating an initial interference matrix Wini and a weight matrix Bini, based on the initial Y and the estimated Uest;c. calculating, using an ML algorithm, trained to determine a non-linear function AH, which correlates an interference matrix (W) and a combining matrix (U), and based on the calculated initial Wini and Bini, an initial optimal Uiopt;d. calculating, using the trained ML algorithm, based on one or more iterations of the determined non-linear function, a final W (Wfin) and a final B (Bfin), to thereby determine a final optimal Ufopt, wherein for each step, the combining matrix is corrected according to a closed-form formula for the optimal U; ande. adjusting beamforming of the receiver unit, based on Ufopt, thereby optimizing the received wireless signal by the receiver unit.

2. The method according to claim 1, wherein the algorithm is executed in the receiver unit.

3. The method according to any one of claims 1-2, executed in real-time.

4. The method according to any one of claims 1-3, wherein the wireless signal comprises a LTE, 5G-NR, mobile 6th generation, Wi-Fi 6-9, Bluetooth, satellite communication signal, GPS, GLONASS, Galileo signal, or any combinations thereof.

5. The method according to any one of claims 1-4, wherein the wireless signal frequency is below 6GHz, or in the range of about 6-100GHz.

6. The method according to any one of claims 1-5, wherein the optimization of the received wireless signal is facilitated at the receiver side only.

7. The method according to any one of claims 1 -6, wherein the optimization of the received wireless signal does not require feedback, a closed-loop-MIMO and / or operation of Physical Upload Share Channels (PUSCH) between the transmitter and the receiver.

8. The method according to any one of claims 1 -7, wherein the receiver unit is a base station, an e / g-nodeB component, a relay, a small cell, a point-to-point transceiver, or any combinations thereof.

9. The method according to any one of claims 1-8, wherein the receiver unit is an end unit.

10. The method according to any one of claims 1-9, wherein the ML algorithm is trained offline, by classifying a collected database comprising a plurality of data patterns, to obtain one or more clusters, wherein each point in a cluster comprises a static interference matrix (Wij) and a control combining matrix (Uj) data set points, having a similar joint influence on the received signal (Yij), wherein the joint influence effect is determined based on a weight matrix Bij; anddetermining, using an ML algorithm, a non-linear function model for the AH, correlating the static interference matrix (W) and the combining matrix (U) and influence on the received signal.

11. The method according to claim 10, wherein the data patterns are created from various interfering areas, said data patterns comprising channel MIMO-wireless interference matrices (AHi, 1=1,... ,L), static interference matrices (Wi), and control combining matrices (Ui).

12. The method according to any one of claims 1-11, wherein the interference comprises:interference signals from general users, interference signals from smart jamming devices, random scattering objects in the wireless channels, substantial multipath effects, active jammers, passive jammers, or any combinations thereof.

13. A system for optimal MIMO-beamforming under interference, to optimize a received wireless signal, the system comprising: a processor configured to execute the method according to any one of claims 1-12.

14. The system according to claim 13, further comprising a receiver unit functionally associated with the processor.

15. An electronic device comprising one or more processors; and memory coupled to the one or more processors, the memory storing one or more programs configured to be executed by the one or more processors, the one or more programs / instructions including instructions for executing the method according to any one of claims 1-12.

16. The electronic device according to claim 15, being a cellular communication device, point to point MIMO wireless communication system and / or advanced WI-FI systems.

17. A non-transitory computer-readable medium storing computer program instructions for executing the method according to any one of claims 1-12.

Images