Energy efficient massive MIMO beamforming using machine learning optimization

Abstract

The present application proposes a novel method and system driven by unsupervised DNN, by developing a precise energy model for each beamforming structure of a massive MIMO system, designing an optimal energy saving hardware configuration and antenna selection for hybrid beamforming and all-digital precoding. The energy model may include power consumption and insertion loss of all components such as a combiner, mixer, power amplifier, etc. The design of the intelligent loss function can provide various tradeoffs between energy consumption and spectral efficiency. The spectrum efficiency, the energy efficiency and the number of active users in the system are considered. The training of the deep unsupervised learning method can use imperfect channel state information. Therefore, the whole process of the proposed DL-based solution is based on the imperfect CSI.

Description

Technical Field

[0001] This disclosure generally relates to the field of wireless communication technology, and more specifically to beamforming technology. Background Technology

[0002] Massive multiple-input multiple-output (mMIMO) technology has revolutionized modern wireless communication, in which base stations (BS) equipped with a large number of antennas transmit signals to multiple users. Hybrid beamforming (HBF) has been proposed to improve energy efficiency and reduce the cost of massive MIMO systems by reducing the number of radio frequency (RF) links in the transmitter [2]. HBF uses a combination of analog and digital precoders consisting of phase shifters and combiners. Typically, three types of hybrid beamforming structures have been proposed: fully connected HBF (FC-HBF), fixed subarray HBF (FSA-HBF), and dynamic subarray HBF (DSA-HBF). In FC-HBF, each RF link is connected to all antennas via a phase shifter, combiner, and power amplifier. In FSA-HBF, each RF link is connected to a subset of antennas, and combiners are removed from the structure to improve implementation cost. To improve the flexibility of FSA-HBF, DSA-HBF has been proposed, in which each antenna is connected to a multiplexer. These multiplexers can dynamically change the connection between the antennas and RF links [7].

[0003] Due to the great success of machine learning (ML), especially deep learning (DL), in many engineering fields, deep neural networks (DNNs) have received much attention in recent years and have been applied to wireless communication systems. Although training DNNs to solve wireless communication problems can be very time-consuming, DNN training can be done offline and only the trained DNN model is used for online decision-making, thus reducing the computational complexity of online computing. Several studies have explored the use of DNNs to solve difficult problems in the physical layer, employing supervised learning, unsupervised learning, and reinforcement learning (RL). On the one hand, in supervised learning, the time spent preparing optimal values ​​(or labels) is not negligible and seems impractical in practice [1]. Moreover, labels must be prepared every time the machine learning model is retrained with a new dataset. On the other hand, reinforcement learning is a promising machine learning approach in which an agent interacts with its environment and makes decisions accordingly [5]. Typically, reinforcement learning does not require a dataset. That is, the agent actively collects data online as it interacts with its environment in a trial-and-error manner. The cost of online data collection can be high due to the large amount of data required. Furthermore, since the action space of HBF is large and continuous, the convergence of reinforcement learning models requires a large number of experiments (i.e., data collection), which complicates HBF in mMIMO systems.

[0004] Currently, the aforementioned technologies still face several challenges. Current DL-based beamforming solutions consider specific HBF structures and are limited to predefined HBF structures. Furthermore, these solutions rely on supervised loss functions, requiring the optimal value to be the objective value, which is computationally expensive and time-consuming. Some unsupervised learning studies have also proposed maximizing the spectral efficiency of predefined beamforming structures without considering hardware constraints or energy efficiency. In addition, one of the most prominent techniques in HBF design is minimizing the Euclidean distance between the required all-digital precoder (FDP) and its hybrid counterpart, which is the objective function used for HBF design [1-6]. Unfortunately, designing FDPs requires this technique, and their performance depends on good channel state information (CSI) acquisition. Therefore, designing HBFs with structures that achieve near-optimal performance is not only computationally expensive but also requires perfect CSI knowledge, an assumption that is difficult to achieve in reality. Invention Summary

[0005] One embodiment of this disclosure includes a method performed by a base station for performing hybrid beamforming or all-digital precoding, for example, in a MIMO system. The method includes: measuring power consumption and insertion loss of one or more components constituting the base station; determining energy consumption based on the power consumption and insertion loss; and detecting the number of user equipment (UEs) communicating with the base station. Further steps include: measuring the spectral efficiency of the base station; comparing the energy consumption, number of UEs, and spectral efficiency with one or more outputs of a trained machine learning model, wherein the trained machine learning model is trained using a smart loss function based at least in part on one or more data related to energy consumption, spectral efficiency, and number of UEs; and creating one or more beamforming structures using multiple antennas based on the trained smart loss function.

[0006] Another embodiment of this disclosure is a method performed by a base station comprising multiple antennas for performing beamforming or all-digital precoding, for example, in a MIMO system. The method includes: measuring the power consumption and insertion loss of one or more components constituting the base station; determining energy consumption based on the power consumption and insertion loss; and detecting the number of UEs communicating with the base station. Further steps of the method include: measuring the spectral efficiency of the base station; training a machine learning model based on a smart loss function, the smart loss function being at least partially based on one or more data related to energy consumption, spectral efficiency, and the number of UEs; and creating one or more beamforming structures using the multiple antennas based on the trained machine learning model.

[0007] Another embodiment includes a network node for performing hybrid beamforming or all-digital precoding. The network node includes: processing circuitry configured to perform any step of the methods described herein based on any network node or base station; and power supply circuitry configured to supply power to the processing circuitry.

[0008] This summary is provided to introduce the selected concepts in a simplified form, which will be further described in detail below. This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to indicate the scope of the claimed subject matter. Attached Figure Description

[0009] To gain a more complete understanding of this disclosure, reference is now made to the following description in conjunction with the accompanying drawings, wherein:

[0010] Figure 1 An embodiment of the base station system disclosed herein is illustrated;

[0011] Figures 2a-2b illustrate embodiments of the base station system of this disclosure;

[0012] Figure 3 An embodiment of the DNN (Deep Neural Network) of this disclosure is shown;

[0013] Figure 4 a-4b illustrates a DNN embodiment of this disclosure;

[0014] Figure 5 An embodiment of the training phase and online phase of this disclosure is illustrated;

[0015] Figure 6 The diagram illustrates the urban layout of the telecommunications system disclosed herein;

[0016] Figure 7 Simulation test results for power consumption and spectral efficiency are shown;

[0017] Figure 8 Simulated test results for energy efficiency and spectral efficiency are shown;

[0018] Figure 9 The simulation test results for some hyperparameters are shown;

[0019] Figure 10 Simulated test results for energy efficiency and spectral efficiency are shown;

[0020] Figure 11 Simulation test results for several hyperparameters are shown;

[0021] Figure 12 The results of the spectral efficiency simulation test are shown;

[0022] Figure 13The results of the spectral efficiency simulation test are shown;

[0023] Figure 14 Simulated test results for power consumption, activated antenna, and energy efficiency are shown.

[0024] Figure 15 A flowchart of an embodiment of the method of this disclosure is shown;

[0025] Figure 16 A flowchart of an embodiment of the method of this disclosure is shown;

[0026] Figure 17 A schematic diagram of an embodiment of the communication system of this disclosure is shown;

[0027] Figure 18 A schematic diagram of a user equipment embodiment of the present disclosure is shown;

[0028] Figure 19 A schematic diagram of an embodiment of the network node of this disclosure is shown;

[0029] Figure 20 A schematic diagram of a host embodiment of this disclosure is shown;

[0030] Figure 21 A schematic diagram of an embodiment of the virtualization environment of this disclosure is shown; and

[0031] Figure 22 A schematic diagram illustrating an embodiment of communication between nodes, hosts, and user equipment according to this disclosure is shown. Detailed Implementation

[0032] Before describing the various embodiments of this disclosure in detail, it should be understood that this disclosure is not limited to the parameters of the specific examples of systems, methods, apparatuses, products, processes, and / or kits, which can, of course, vary. Therefore, while certain embodiments of this disclosure will be described in detail with reference to specific configurations, parameters, components, elements, etc., these descriptions are illustrative only and should not be construed as limiting the scope of the claimed embodiments. Furthermore, the terminology used herein is intended to describe embodiments and is not necessarily intended to limit the scope of the claimed embodiments.

[0033] Certain aspects of this disclosure and the embodiments described herein can provide solutions to the above and other challenges in the art as described herein.

[0034] For example, consider a time-division duplex (TDD) downlink massive MIMO system where one base station (BS) serves a group of users. One objective of this disclosure is to maximize energy efficiency and achieve a balance between energy efficiency and spectral efficiency by designing a novel deep unsupervised learning algorithm that can find different beamforming structures for both all-digital precoder (FDP) and hybrid beamforming (HBF) structures. A loss function for training the unsupervised learning algorithm can be designed based on an accurate transmitter energy model to improve energy efficiency. For the FDP case, the embodiment includes learning a method for performing antenna selection, while for HBF, it reduces output power and the number of radio frequency (RF) chains used.

[0035] Some proposed algorithms can perform antenna selection and hardware configuration by considering the power consumption and insertion loss of all components in each beamforming (BF) structure. To satisfy the connectivity constraints of different BF structures (which have discrete characteristics), the described unsupervised learning algorithm utilizes the Gumbel-Sigmoid method, inspired by Gumbel-Softmax. The Gumbel-Sigmoid algorithm is designed to consider the constraints of all components involved in the BF connectivity.

[0036] In the context of large-scale MIMO beamforming based on deep learning, the described embodiment is the first to use imperfect channel state information (CSI) to train a deep neural network (DNN), not only as input to the DNN, but also for computing an unsupervised loss function.

[0037] The embodiments described herein include novel algorithms driven by unsupervised deep neural networks (DNNs) that design optimal energy-efficient hardware configurations and antenna selections for HBF and FDP by developing accurate energy models for each beamforming structure of a massive MIMO system. This energy model includes the power consumption and insertion loss of all components such as combiners, mixers, and power amplifiers.

[0038] The embodiments also provide the design of intelligent loss functions that can provide various trade-offs between energy consumption and spectral efficiency. The embodiments may take into account spectral efficiency, energy efficiency, and the number of active users in the system.

[0039] The embodiments also include the use of imperfect channel state information during the training process of the deep unsupervised learning method. Therefore, the entire process of the proposed DL-based solution can be based on imperfect CSI.

[0040] Certain embodiments may offer one or more of the following technical advantages. The proposed deep unsupervised learning algorithm is flexible and adaptable to various hardware configurations, such as hybrid and all-digital architectures. The proposed algorithm can be trained efficiently with only minor modifications depending on the beamforming configuration. Typically, the output layer of the DNN varies with the beamforming configuration. The loss function for each beamforming architecture is designed to include three terms weighted by some hyperparameters:

[0041] Loss=-SE+γEC+ζAS

[0042] Equation 1

[0043] Here, SE represents the achieved spectral efficiency, EC represents the energy consumption, and AS represents the adaptive antenna selection based on the desired spectral efficiency. Hyperparameters γ and ζ are used to control the weights of each term to obtain the SE-EE tradeoff.

[0044] Because the proposed algorithm is based on an accurate energy model, the DNN model can comprehensively consider hardware configuration and its energy consumption to design beamforming solutions. Furthermore, the proposed loss function not only reflects the goal of maximizing energy efficiency but also supports various trade-offs between spectral efficiency and energy consumption.

[0045] Some proposed unsupervised DNNs can be trained using only noisy CSI. Therefore, the entire process (including the training and evaluation phases) can be performed using CSI obtained during normal base station operation.

[0046] Some embodiments contemplated herein will now be described more fully with reference to the accompanying drawings. These embodiments are provided by way of example only and are intended to convey the scope of the subject matter to those skilled in the art.

[0047] System Model

[0048] You can refer to this. Figure 1 The illustration shows a system embodiment. System 100 is a downlink massive MIMO system, in which one BS140 sends a signal to N... U Each user transmits 180 units using a single antenna. The BS140 is equipped with N... T 120 and N antennas RF Each RF link 110. A digital precoder (DP) 106 processes the received data 102 in baseband 105, and then outputs the signal through the RF links 110. Each RF link 110 consists of a digital-to-analog converter (DAC), a low-pass filter (LPF), a local oscillator (LO), and a mixer. The connection between the RF links 110 and the antenna 120 defines an analog precoder (AP) 116, which can be implemented using phase shifters, switches, and combiners. Figure 1An embodiment of a large-scale MIMO system model architecture is shown, in which a transmitter BS140 uses HBF to serve a group of users 180.

[0049] The signal received by each user u is given by the following formula:

[0050]

[0051] in It is the channel vector of user u. It is the sending symbol for all users, and η is a variable with a mean of 0 and a variance of σ. 2 Additive white Gaussian noise.

[0052] HBF includes DP And APA = P q ×Ω, where This is the coefficient matrix of the q-bit phase shifter connecting antenna 120 and RF link 110. The coefficients of the q-bit phase shifter connecting the nth antenna 120 and the mth RF link 110 can be given by the following formula:

[0053]

[0054] The matrix Ω defines the connection state between the antenna and the RF link. It is a binary matrix whose (n, m)th element is 1 if and only if the nth antenna 120 is connected to the mth RF link 110.

[0055] The achievable spectral efficiency (SE) of a massive MIMO system is given by the following equation:

[0056]

[0057]

[0058] Where SINR(A,w) u The ratio of the signal received by user u to the interference plus noise is given by the following formula:

[0059]

[0060] One objective is to maximize the energy efficiency (EE) of a massive MIMO system. EE is defined as the ratio of SE to power consumption P. HBF The ratio between them will be further modeled and equationd below. The mathematical problem is equationd as follows:

[0061]

[0062] Where P TX This is the power budget for BS.

[0063] Based on this equation, a baseline solution (an approximate solution for the optimal FDP and HBF architectures) can be obtained for comparison with the proposed deep unsupervised learning solution. It is assumed that time-division duplexing is used, where the estimated channel for the uplink can be used for the downlink. To more closely approximate reality, it is assumed that the wireless channel is noisy, i.e.:

[0064]

[0065] Where H is the actual wireless channel matrix, and ∈ is the added Gaussian white noise. The parameter β∈[0,1] is a hyperparameter used in the model to study the impact of noise on the performance of the proposed solution.

[0066] Energy Model

[0067] To optimize the energy dissipation (EE), the power consumption of the proposed massive MIMO system should be defined. This can be done based on the general assumption that components of the same type have the same input / output interfaces; that is, their inputs and outputs are connected to the same type and number of components. This assumption generally holds true because it simplifies the concept of a general-purpose circuit. Furthermore, there is no need for differentiated designs for different antennas, as manufacturing circuits for specific applications is costly.

[0068] To better represent each HBF structure, a general template is used, as shown in Figures 2a and 2b. Figure 2a illustrates an embodiment of HBF structure 300a. Figure 2b illustrates an embodiment of FDP structure 300b. Both HBF structure 300a and FDP structure 300b include one or more RF links 310a / b leading to one or more connections 320a / b, then to one or more power amplifiers 343a / b, and further to one or more antennas 345a / b. The different locations for measuring / detecting / monitoring power are marked in both structures as follows. Total baseband power output; The input power of the power amplifier; and Output power of the power amplifier.

[0069] As shown in Figure 2a, given antenna 345a, it is connected to a configuration with c∈{1,…,N} RF A combiner 340a has ψ inputs. Each input of the combiner 340a is connected to the output of a phase shifter 335a. Each phase shifter 335a is then connected to an RF link 310a via a switch 330a. The number of switches 330a is ψ∈{1,…,N}. RF The tuple (ψ, c) is defined to fully characterize the analog precoder structure. The HBF structure 300a can be applied to the three possible HBF structures discussed earlier:

[0070] • (NRF, NRF) for the FC-HBF structure. In the FC-HBF structure, all switches are connected (i.e., ψ = N). RF The outputs of all phase shifters are combined before each antenna (i.e., c = N). RF ).

[0071] (NRF,1) For the DSA-HBF structure. In DSA-HBF, only one switch can be connected at each time slot, so c = 1, while all switches can potentially be connected, so ψ = N. RF It is important to note that this switch configuration functions similarly to a multiplexer. Therefore, in a practical system, the switch is replaced by a ψ×1 multiplexer.

[0072] (1,1) For the FSA-HBF structure. In FSA-HBF, each RF link is connected to only one antenna (i.e., c = 1), and the connection is fixed (i.e., ψ = 1).

[0073] Table 1 compares the hardware complexity of different beamforming technologies.

[0074]

[0075] Table 1: Comparison of Hardware Complexity

[0076] The energy consumption of each component will be described below, along with a list of the latest and most advanced hardware solutions. Furthermore, since the operating frequency is assumed to be 28 GHz, components operating in the 20-40 GHz frequency range are also included.

[0077] A component consists of symbols <c>The identifier corresponds to one element in the sets D, L, M, LO, Ψ, Φ, C, and A. The correspondence between components and their symbols is defined in Table 2. We will use IL <c>< / c> Expressed as the insertion loss of a passive component, P <c>< / c> (x) represents an active component. <c>The average power dissipated depends on a set of parameters x defined in Table 2. Note that the power dissipated by the wires is ignored, and when c = 1, a synthesizer (i.e., IL) is not required. C (1) = 0dB). Similarly, for a switch, if ψ = c, it means that all connections have been established, and when ψ = 1, the switch behaves like a wire (i.e., IL). Ψ (c) = IL Ψ (1) = 0dB).

[0078] The components to be discussed include the RF (radio frequency) front end, passive components, digital-to-analog converter, and low-pass filter in the transceiver (TX).

[0079] RF Front End: The RF front end refers to the circuitry between the antenna and the DAC. As shown in Figure 2b, for the FDP, it includes a low-pass filter (LPF), mixer, local oscillator (LO), switch, and power amplifier (PA). On the other hand, in Figure 2a, the HBF, in addition to the components described in the FDP, uses a network consisting of phase shifters, splitters, and combiners. The mixer, combiner, and PS are considered passive devices that introduce IL into the circuitry.

[0080]

[0081] Table 2: Energy Model Symbols and Parameters

[0082] Passive components: Mixers, combiners, and PS are all assumed to be passive devices that introduce IL. The insertion loss of phase shifters and combiners plays a crucial role in designing energy-efficient HBFs, especially for FC-HBFs, where all RF links are connected to all antennas via phase shifters and combiners. For DSA-HBFs, switches dynamically alter the connections between RF links and antennas to improve structural flexibility.

[0083] For simplicity, we assume all ILs are linearly proportional. Now, assume the total baseband output power is... The input power of the PA of the nth antenna can be written in the following form for all HBF structures:

[0084]

[0085] Among them IL Φ (q) represents the insertion loss of the PS at q-bit resolution. In FC-HBF, all RF links are connected to the antenna, and (ψ,c) = (N RF N RF And IL Ψ (c) = 0 dB, then the above formula can be rewritten as:

[0086]

[0087] Similarly, for structures with (ψ,c)=(N) RF ,1) and the connection matrix Ω SA The DSA-HBF, and the structure (ψ,c)=(1,1) and the connection matrix Ω DSA For the FSA-HBF, the input power of the DSA-HBF can be written as:

[0088]

[0089] and

[0090]

[0091] Where N S =N T / N RF This indicates the size of the connected subarray. Similarly, for the FDP, as shown in Figure 2b, the input power of the PA can be obtained as:

[0092]

[0093] Based on the beamforming structure given above, assuming The DC power absorbed by the nth PA can be written as:

[0094]

[0095] Where α is the power-added efficiency (PAE) of the LPA. It is the transmit power of the nth antenna, where

[0096] It should also be noted that, due to the total power constraint, the output power of all antennas is not necessarily equal, and the total transmit power is limited to P. TX .

[0097] Digital-to-Analog Converter (DAC): The DAC is one of the most power-consuming components in wireless applications. The power consumption (PD) of a DAC is determined by the converter's sampling frequency (f / Hz). s ) and quality factor (FOM) D It is a linear function of resolution bits (b) and varies with resolution bits (b) D It grows exponentially with the increase of ), such as: P D =FoM D ×f s ×2 bD The sampling frequency range for ultra-wideband applications is 0.5-1 GHz. In terms of the required signal-to-noise ratio (SQNR), FDP requires 2 fewer bits than HBF.

[0098] The low-pass filter in TX: The DAC output needs an analog LPF to suppress spectral mirroring and maintain out-of-band emission limits. For a cutoff frequency of f... c m-order active LPF, FoM L This is the power consumption per pole per hertz. The power consumption of an LPF is determined by P. L =FoM L ×f c ×m′ is given.

[0099] Now, putting them together, the total power consumed by a given beamforming structure can be written in the following form:

[0100]

[0101] Where P LO This represents the power consumed by the mixer from the LO. It should be noted that the above equation represents total power consumption, denoted by the symbol P. HBF Let HBF be an expression, and for FDP, when N RF =N T When, use P FDP The power consumption of different HBF structures is almost the same due to the low insertion loss of passive components before the PA, based on the power consumption of passive components such as phase shifters and combiners. However, in terms of hardware complexity and cost, as shown in Table 1, subarray HBF is more efficient than FC-HBF.

[0102] Based on the previously defined connection matrix Ω, the above equation can be written more generally as follows:

[0103]

[0104] Where ||.||0 represents the cardinality of the vector, and Ω∈{Ω FD ,Ω FC ,Ω SA As can be seen, both SE and EE depend on the matrix Ω, which defines the connections between the RF link and the antenna. More connections result in a higher SE due to increased beamforming flexibility. However, in FDP, each connection corresponds to the use of one RF link, while in HBF, each connection corresponds to one PS and one combiner, leading to higher energy consumption.

[0105] The proposed energy-efficient beamforming driven by deep unsupervised learning

[0106] The following describes some implementations of unsupervised learning solutions for designing antenna selection, efficient HBF, and FDP. The proposed algorithm consists of two phases: (i) the training phase and (ii) the online phase. First, the DNN architecture is described.

[0107] Deep neural network architecture:

[0108] Figure 3 The figure illustrates a proposed DNN. core The implementation is similar to the proposed HBF and FDP. Since each BF structure requires a different DNN output, the similar DNN parts of these two structures can be described first, referred to as "DNN". core ",like Figure 3 As shown. DNN core Contains two convolutional layers (CL) 16@N T ×N U Where 16 represents the number of channels (or filters), and N T ×N U This is the dimension of each channel, followed by 1CL 8@N T ×N U All line units (CLs) have a kernel size of 3×3. Following the CLs are two fully connected layers (FLs), each with 512 neurons. All layers are followed by the Leaky ReLU activation function. Leaky ReLU is a ReLU-based activation function. Leaky ReLU has a smaller slope for negative values, rather than a flat slope.

[0109] Batch normalization is used after each layer to avoid overfitting. The input to the DNN is a noise channel matrix. To improve representation learning, we first target the channels. Perform normalization, and then The real and imaginary parts are denoted as follows: and It is separated into two channels in the first CL.

[0110] HBF output layer

[0111] Figure 4 Figures a and 4b illustrate embodiments of the proposed DNN architecture for (a) hybrid beamforming and (b) all-digital precoder.

[0112] like Figure 4 As shown in figure a, the output of the last FL is divided into four parallel fully connected layers. Their depth depends on the desired output dimension. The size of the first and second parallel layers is N. RF ×N U This generates the real and imaginary parts of the dynamic programming (DP). The output of the third parallel layer generates the dynamic programming (AP), therefore its dimension is N. RF ×N T The AP's output can also be adapted to different phase shifter resolutions.

[0113] Size N RF ×N T The fourth layer design matrix Ω HB As mentioned above, Ω HB ∈{Ω FC ,Ω FSA ,Ω DSA The matrix must be a binary matrix. Typically, this binary constraint requires the use of the Sigmoid function during training, followed by rounding techniques to convert real values ​​to binary values ​​during the online phase. However, the applicant found this approach unsatisfactory for unsupervised learning because the SE measured during training can differ significantly from the actual SE measured during testing. This is because there are no labels in unsupervised learning methods, so the DNN's output is not saturated with binary values. To address this issue, inspired by the Gumbel-Softmax estimator, a differentiable approximation called Gumbel-Sigmoid is proposed to be used during training. The Gumbel-Softmax approximation is a technique that allows sampling from a categorical distribution during the neural network's forward propagation, combining reparameterization tricks and smooth relaxation. Therefore, a categorical binary distribution can be used to represent the connection between the RF link and the antenna. Thus, π is defined... n,m Let N be the probability that antenna n is connected to RF link m, and thus we can construct N T ×N RF The matrix Π corresponds to the probabilistic states between antenna n and RF link m. The Gumbel-Softmax function G(Π) applied to each element of matrix Π can be defined as follows:

[0114]

[0115] Because a binary classification distribution is assumed, the equation is:

[0116]

[0117] and

[0118]

[0119] Where Ω HB This is the output of the DNN, where g and g' are independent samples following a Gumbel distribution with zero mean and unit variance. It's important to note that when a matrix is ​​used as input, the exp(·) and log(·) functions are applied element-wise. The parameter τ is called the Gumbel temperature. As τ→0, G(Π) tends towards a categorical distribution, while as τ→∞, it converges to a uniform distribution. Therefore, there is a trade-off between a smaller temperature (sample vectors are closer to a one-hot distribution but have a larger gradient variance) and a larger temperature (samples are more uniform but have a smaller gradient variance). Thus, τ is considered a hyperparameter to be optimized in our implementation.

[0120] FDP output layer

[0121] Figure 4 b illustrates a proposed architecture for the FDP. The output layer can be divided into three parallel layers. The first two layers are dedicated to the real and imaginary parts of the FDP, with a dimension of N. T ×N U The third layer, similar to HBF, is used to design the antenna selection vector (ω), where π is assumed. ′ n Let be the probability of connecting to antenna index n, then ω = G(π'), where Final Ω FD =diag(ω).

[0122] Training phase: Unsupervised learning

[0123] Figure 5 The training phase of the proposed DNN is shown. For example... Figure 5 As shown, in one embodiment of the proposed algorithm, it is assumed that during the training phase, the BS implementing the algorithm does not transmit any data, but only measures the environment and stores data samples. In this case, due to unsupervised learning, the data samples can constitute a noise channel matrix without a target (or label). The noise term includes coefficients β and calibration error ∈, as shown in Equation 8. The coefficients β are used to control the noise power, thus helping to study the impact of noise on the proposed DNN unsupervised learning method. To make the system model more realistic, the noise channel model is used even during the training phase to compute the loss function. This makes the proposed model more realistic compared to existing state-of-the-art models that primarily assume a perfect channel during training.

[0124] High-efficiency hybrid beamforming (E-HBF-Net)

[0125] Initially, one proposed algorithm could start with an HBF structure called E-HBF-Net, where all RF links are connected to all antennas via PS. However, to design an efficient HBF structure, the proposed algorithm requires each connection (N) to be optimized. T ×N RF A programmable switch is used to find the optimal matrix (Ω) that maximizes EE. HB ).like Figure 5 As shown in the training phase, DNNs are being jointly designed (i) (ii) Having a regression task And (iii) by employing the proposed Gumbel Sigmoid function in the RF link and antenna (Ω) HB The proposed DNN aims not only to design the HBF to maximize SE, but also to design the connection matrix to improve EE. Furthermore, the proposed DNN is adaptive to the number of active users; when the number of active users is low, the proposed DNN intelligently shuts down some antennas because these antennas are no longer needed. Therefore, it reduces energy consumption.

[0126] Nevertheless, the unsupervised loss function used to train the DNN uses three terms, which are written as described in Equation 1, with reference to hybrid beamforming.

[0127] Loss HBF =-SE+γEC+ζAS

[0128] Equation 20

[0129] Where SE represents the achieved spectral efficiency, EC represents the energy consumption, and AS represents the adaptive antenna selection based on the desired spectral efficiency. Hyperparameters γ and ζ are used to control the weight of each term to obtain the SE-EE tradeoff.

[0130] Maximize spectral efficiency (SE)

[0131] The first term of the loss function corresponds to maximizing SE by optimizing AP and DP. Therefore, the first term SE is given by the negative of SE, as follows:

[0132]

[0133] in here, This represents the DNN output of the phase shifter. This represents the connection matrix output by the DNN after using the "Gumbel Sigmoid" function to compute SE in a noisy channel. Equation 21 can be compared with Equation 6 above. Equation 6 deals with maximizing EE, defined as SE (see Equation 21) divided by power consumption. The term SE in the loss function is used to maximize SE under the assumption of total power constraint at BS. Therefore, to satisfy the power constraint, the power can be normalized to... However, due to consideration The power is very low; this power normalization ensures that each PA has a constant power consumption, independent of the connection matrix. The transmission power can be renormalized to... The function.

[0134]

[0135] Minimize energy consumption (EC)

[0136] The second term also corresponds to the connection matrix in DNN design. This term is introduced to add a penalty to the total loss function to reduce energy consumption. It is given as follows:

[0137] EC = γP HBF

[0138] Equation 23

[0139] Among them, P HBF The total energy consumption depends on the number of power supplies (PS), combiners, and power amplifiers (PAs) used. therefore, It affects both SE (Sequence Energy) and EE (Energy Efficiency). The parameter γ in the above equation is a regularization coefficient used to balance SE and EE. It is a hyperparameter that can be optimized to improve the flexibility of BF (Browser Flow) design. It can be seen that when γ→∞, the DNN ignores energy consumption and focuses on maximizing SE for maximum flexibility. On the other hand, when γ→0, the DNN sacrifices SE to minimize energy consumption. Simulation results show the impact of different γ values.

[0140] To calculate P HBF We need to calculate the power consumed, so we first need to know the DC power consumed by the PA. Therefore, we need to know the input and output power of the PA. The output power of the PA in the nth antenna can be written as:

[0141]

[0142] in and The AP and DP are designed by a DNN. However, obtaining the input power of the PA is challenging because the connection matrix is ​​designed by the DNN and is not in the well-known FC-HBF and SA-HBF forms. Therefore, the input power cannot be calculated directly. It must depend on the connection matrix. This matrix determines the power divider after each RF link and the power combiner before each antenna. Therefore, the input power of the PA can be written as:

[0143]

[0144] in express The nth line, We should determine the number of RF links used to calculate the total power consumption of the RF links. However, finding the number of RF links requires using a cardinality function, which is not a differentiable operation, causing the backpropagation algorithm to fail. Therefore, the expectation for all antennas is defined as follows:

[0145]

[0146] As mentioned above, It is the output of the Gumbel-Sigmoid function.

[0147] Adaptive Antenna Selection (AS)

[0148] The third term AS is given as follows:

[0149]

[0150] Where R desire This is the predefined expected average SE value for all users. The first two terms in the loss function balance SE and energy consumption by designing the precoder without considering the number of active users. In an mMIMO system, each user needs a predefined SE threshold R. desire In this case, the DNN should no longer focus on maximizing SE, but rather on minimizing power consumption. Therefore, the third term is defined as the average SE versus R. desire The difference between them, where the average SE depends on the number of active users. Due to the effect of this factor, SE is forced to approach R. desire This will not exceed the limit, and some unnecessary antennas (transmit power) can be turned off to reduce energy consumption (according to the second term EC). Therefore, this term guarantees that the minimum power consumption is met to satisfy the target average rate R. desire .

[0151] R desire The value of is fixed and predefined by the operator, where higher values ​​result in higher power consumption, while lower values ​​reduce the number of available antennas. For example, consider a scenario with four users, an average rate of 6 b / s / Hz per user, and an optimal SE of 24 b / s / Hz; another scenario with two users, an average rate of 9 b / s / Hz per user, and an optimal SE of 18 b / s / Hz. Assume R... desire =7b / s / Hz. The first scenario has less impact on the DNN solution than the second scenario because AS=1 in the first scenario and AS=4 in the second. Therefore, in the second scenario with two users, the DNN must reduce the number of transmit antennas to reduce the average rate per user from 9b / s / Hz to close to R. desire = 7b / s / Hz.

[0152] All-digital precoder (E-FDP-Net)

[0153] FDP's DNN provides a pre-encoder and the vector used for antenna selection For FDP, the unsupervised loss function for training the DNN is similar to that for HBF, calculated according to Equation 20 (Loss = -SE + γ′EC + ζAS) by finding the SE, EC, and AS terms used for FDP. Mathematically defined as follows:

[0154]

[0155] Among them, the SE of FDP is composed of Given that the EC of FDP is given by P FDP Given, and with a new hyperparameter γ ′ The AS of FDP depends on SE.

[0156] The proposed deep unsupervised learning algorithm comprises a deep neural network with wireless channels. As input and using the loss function or (Depending on the BF architecture) Training is performed as shown in Figure 2.

[0157] Online phase: Data transmission

[0158] After the training phase ends, when the DNN is ready for inference, the online phase begins, such as... Figure 5 As shown. In the online phase, the input to the DNN consists only of the noise channel matrix. Given. In the online phase, similar to the training phase, HBF... and as well as It can be used without further processing. However, since the connection matrix is ​​binary, i.e., in HBF... and FDP They require binary quantization. For this, element-wise rounding functions can be used for each element of these matrices. As follows: For HBF, for

[0159] FDP, and The output power of E-HBF-Net can be achieved through... Obtained, and for E-FDP-Net, through

[0160] Simulation results

[0161] The applicant conducted multiple simulations using the method described herein. The channel simulation results are as follows.

[0162] Some embodiments of deep unsupervised learning solutions require a wireless channel as input and a BF matrix as output. One solution proposed in this disclosure can be evaluated using a real ray-traced channel model called "deepMIMO". This dataset contains different large-scale MIMO scenarios, and simulations were implemented using the "O1-28 GHz" scenario. The wireless channel was generated by applying a ray tracing method to a 3D model of an urban environment. The "O1-28 GHz" scenario utilizes multiple user locations randomly generated in two orthogonal streets that intersect in the middle of the area and are surrounded by buildings, such as... Figure 6 As shown. Figure 6 The O1-28 GHz scene of the deepMIMO dataset is shown.

[0163] Figure 7 The results show the relationship between power consumption (W) and achievable spectral efficiency (b / s / Hz). The figure illustrates the trade-offs between SE and EE when the hyperparameters γ and γ' are varied. The figure also shows a comparison of some proposed solutions (E-FDP-Net and E-HBF-Net) with optimal and benchmark conventional solutions including FDP, FC-HBF (MO-AltMin), DSA-HBF, and FSA-HBF.

[0164] Figure 8 This shows the trade-off between SE and EE when the hyperparameter γ' is changed for the FDP scenario. It can be seen that as EE increases, SE decreases, and vice versa. The crossover point is located at approximately γ' = 4.

[0165] Figure 9 This displays the connection between the RF link and antenna for FDP under different hyperparameter γ′ values. Gray squares represent connections (binary value 1), and white squares represent no connections (binary value 0).

[0166] Figure 10 This shows the trade-off between SE and EE when the hyperparameter γ is changed for the HBF scenario. It can be seen that as EE increases, SE decreases, and vice versa. The crossover point is located at approximately γ = 0.7.

[0167] Figure 11 This displays the connection status between the RF link and antenna for HBF under different hyperparameter γ values. Gray squares represent connections (binary value 1), and white squares represent no connections (binary value 0).

[0168] Figure 12 The results show the SE achieved by the proposed solutions (E-FDP-Net and E-HBF-Net) compared to other benchmark methods (FDP, MO-AltMin, PE-AltMin) when the noise parameter β is varied. A higher β value indicates a larger noise variance.

[0169] Figure 13 The proposed solutions (E-FDP-Net and E-HBF-Net) are shown to achieve SE for different noise variances compared to benchmark solutions (FDP, MO-AltMin, PE-AltMin, FC-HBF-Net, MO-AltMin, DSA-HBF-Net and FSA-HBF-Net).

[0170] Figure 14 The display shows the change in the number of active users N. U and SE target R target At that time, FDP achieves the number of active antennas and EE respectively.

[0171] Additional Examples

[0172] Figure 15 Another possible embodiment of this disclosure is illustrated. Method 1500 includes a method performed by a base station for performing hybrid beamforming or all-digital precoding, for example, in a MIMO system. Step 1510 is to measure the power consumption and insertion loss of one or more components constituting the base station. Step 1520 is to determine energy consumption based on the power consumption and insertion loss. Step 1530 is to detect the number of UEs communicating with the base station. Step 1540 is to measure the spectral efficiency of the base station. Step 1550 is to compare the energy consumption, number of UEs, and spectral efficiency with one or more outputs of a trained machine learning model, wherein the trained machine learning model is trained using a smart loss function that is at least partially based on one or more data related to the energy consumption, spectral efficiency, and number of UEs. Step 1560 is to create one or more beamforming structures using multiple antennas based on the trained smart loss function. The creation of the beamforming structures can be accomplished by switching activities, for example, establishing a connection between an RF link and antennas, thereby generating different beamforming structures at the base station.

[0173] Figure 16 Another possible method embodiment under this embodiment is shown. Method 1700 is a method performed by a base station including multiple antennas for, for example, performing beamforming or all-digital precoding in a MIMO system. Step 1710 is to measure the power consumption and insertion loss of one or more components constituting the base station. Step 1720 is to determine energy consumption based on power consumption and insertion loss. Step 1730 is to detect the number of UEs communicating with the base station. Step 1740 is to measure the spectral efficiency of the base station. Step 1750 is to train a machine learning model based on a smart loss function, which is at least partially based on one or more data related to energy consumption, spectral efficiency, and the number of UEs. Step 1760 is to create one or more beamforming structures using multiple antennas based on the trained machine learning model.

[0174] Figure 17 An example of a communication system 2100 according to some embodiments is shown. In this example, the communication system 2100 includes a telecommunications network 2102, which includes an access network 2104 and a core network 2106, such as a RAN, which includes one or more core network nodes 2108. The access network 2104 includes one or more access network nodes, such as network nodes 2110a and 2110b (one or more of which may be collectively referred to as network node 2110), or any other similar 3GPP access node or non-3GPP access point. Network node 2110 facilitates direct or indirect connections for UEs, such as connecting UEs 2112a, 2112b, 2112c, and 2112d (one or more of which may be collectively referred to as UE 2112) to the core network 2106 via one or more radio connections.

[0175] Examples of wireless communication via a wireless connection include sending and / or receiving wireless signals using electromagnetic waves, radio waves, infrared waves, and / or other signal types suitable for transmitting information without the use of wires, cables, or other physical conductors. Furthermore, in various embodiments, communication system 1100 may include any number of wired or wireless networks, network nodes, user equipment (UEs), and / or any other components or systems that can facilitate or participate in communication of data and / or signals via wired or wireless connections. Communication system 2100 may include and / or interact with any type of communication, telecommunications, data, cellular, radio network, and / or other similar types of systems.

[0176] UE 2112 can be any of a variety of communication devices, including wireless devices that are arranged, configured, and / or operable for wireless communication with network node 2110 and other communication devices. Similarly, network node 2110 is arranged, capable of, configured, and / or operable for communicating directly or indirectly with UE 2112 and / or with other network nodes or devices in telecommunication network 2102 to achieve and / or provide network access, such as wireless network access, and / or perform other functions, such as management in telecommunication network 2102.

[0177] In the described example, core network 2106 connects network node 2110 to one or more hosts, such as host 2116. These connections can be direct or indirect, via one or more intermediate networks or devices. In other examples, network nodes can be directly coupled to hosts. Core network 2106 includes one or more core network nodes (e.g., core network node 2108) consisting of hardware and software components. The characteristics of these components may be substantially similar to those described for user equipment (UE), network nodes, and / or hosts, and therefore their description generally applies to the corresponding components of core network node 2108. Example core network nodes include one or more of the following functions: Mobile Switching Center (MSC), Mobility Management Entity (MME), Home Subscriber Server (HSS), Access and Mobility Management Function (AMF), Session Management Function (SMF), Authentication Server Function (AUSF), Subscriber Identity Decryption Function (SIDF), Unified Data Management (UDM), Secure Edge Protection Agent (SEPP), Network Exposure Function (NEF), and / or User Plane Function (UPF).

[0178] Host 2116 may be owned or controlled by a service provider other than the operator or provider of access network 2104 and / or telecommunications network 2102, and may be operated by the service provider or operated on its behalf. Host 2116 may host various applications to provide one or more services. Examples of such applications include live and pre-recorded audio / video content, data collection services (e.g., retrieving and compiling data on various environmental conditions detected by multiple UEs), analytics functions, social media, functions for controlling or otherwise interacting with remote devices, alarm and monitoring center functions, or any other such functions performed by a server.

[0179] Overall, Figure 17 The communication system 2100 enables connections between the UE, network nodes, and hosts. In this sense, the communication system can be configured to operate according to predefined rules or procedures, such as specific standards, including but not limited to: Global System for Mobile Communications (GSM), Universal Mobile Telecommunications System (UMTS), Long Term Evolution (LTE), and / or other suitable 2G, 3G, 4G, 5G standards or any applicable future generation standard (e.g., 6G), Wireless Local Area Network (WLAN) standards such as the IEEE 802.11 standard (WiFi), and / or any other suitable wireless communication standards such as Global Microwave Access Interoperability (WiMax), Bluetooth, Z-Wave, Near Field Communication (NFC) ZigBee, LiFi, and / or any Low Power Consumption Wide Area Network (LPWAN) standards, such as LoRa and Sigfox.

[0180] In some examples, telecommunications network 2102 is a cellular network implementing 3GPP normalization features. Therefore, telecommunications network 2102 can support network slicing to provide different logical networks to different devices connected to it. For example, telecommunications network 2102 can provide ultra-reliable low-latency communication (URLLC) services to some UEs while providing enhanced mobile broadband (eMBB) services to other UEs and / or massive machine-type communication (mMTC) / massive IoT services to even more UEs.

[0181] In some examples, UE 2112 is configured to send and / or receive information without direct human-machine interaction. For example, the UE may be designed to send information to access network 2104 according to a predetermined schedule, triggered by internal or external events, or in response to requests from access network 2104. Furthermore, the UE may be configured to operate in single RAT, multiple RAT, or multiple standard modes. For example, the UE may operate with any combination of Wi-Fi, NR (New Radio), and LTE, i.e., configured as multi-radio dual connectivity (MR-DC), such as E-UTRAN (Evolved UMTS Terrestrial Radio Access Network) New Radio-Dual Connectivity (EN-DC).

[0182] In this example, hub 2114 communicates with access network 2104 to facilitate indirect communication between one or more UEs (e.g., UEs 2112c and / or 2112d) and network nodes (e.g., network node 2110b). In some examples, hub 2114 may be a controller, router, content source, and analyzer, or any other communication device described herein with respect to a UE. For example, hub 2114 may be a broadband router for enabling UE access to core network 2106. As another example, hub 2114 may be a controller for sending commands or instructions to one or more actuators in the UE. Commands or instructions may be received from the UE, network node 2110, or via executable code, scripts, processes, or other instructions within hub 2114. As another example, hub 2114 may be a data collector that acts as a temporary storage for UE data and, in some embodiments, may perform data analysis or other processing. As yet another example, hub 2114 may be a content source. For example, for a UE acting as a VR headset, display, speaker, or other media delivery device, hub 2114 can retrieve VR resources, video, audio, or other media, or data related to sensing information, from network nodes, and then provide it to the UE directly, after performing local processing, and / or after adding other local content. In yet another example, hub 2114 acts as a proxy server or coordinator for the UE, especially when one or more UEs are low-power IoT devices.

[0183] Hub 2114 can maintain a constant / persistent or intermittent connection with network node 2110b. Hub 2114 can also allow different communication schemes and / or scheduling between hub 2114 and UEs (e.g., UEs 2112c and / or 2112d) and between hub 2114 and core network 2106. In other examples, hub 2114 is connected to core network 2106 and / or one or more user equipments (UEs) via a wired connection. Furthermore, hub 2114 can be configured to connect to an M2M service provider via access network 1104 and / or to another user equipment (UE) via a direct connection. In some cases, user equipment (UE) can establish a wireless connection with network node 2110 while still being connected via hub 2114 via a wired or wireless connection. In some embodiments, hub 2114 can be a dedicated hub, i.e., its primary function is to route communication from / to network node 2110b. In other embodiments, hub 2114 may be a non-dedicated hub, i.e. a device capable of routing communication between user equipment (UE) and network node 2110b, but it may also serve as the starting and / or ending point of communication for certain data channels.

[0184] Figure 18 A UE 2200 according to some embodiments is illustrated. As used herein, "UE" refers to a device capable of, configured, arranged, and / or operated for wireless communication with network nodes and / or other UEs. Examples of UEs include, but are not limited to: smartphones, mobile phones, cellular phones, Voice over IP (VoIP) phones, wireless local loop phones, desktop computers, personal digital assistants (PDAs), wireless cameras, game consoles or devices, music storage devices, playback devices, wearable terminal devices, wireless endpoints, mobile stations, tablets, laptops, laptop embedded devices (LEEs), laptop mounted devices (LMEs), smart devices, wireless client devices (CPEs), in-vehicle or in-vehicle embedded / integrated wireless devices, etc. Other examples include any UE identified by the 3rd Generation Partnership Project (3GPP), including Narrowband Internet of Things (NB-IoT) UEs, Machine Type Communication (MTC) UEs, and / or Enhanced MTC (eMTC) UEs.

[0185] The UE can support device-to-device (D2D) communication, for example, by implementing 3GPP standards for sidechain communication, Dedicated Short Range Communication (DSRC), vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), or vehicle-to-everything (V2X). In other examples, the UE may not necessarily be a user in the sense of a human user who owns and / or operates the associated equipment. Instead, the UE may represent equipment intended to be sold to or operated by a human user, but which may not be associated with a particular human user, or may not initially be associated with a particular human user (e.g., a smart irrigation controller). Alternatively, the UE may represent equipment not intended to be sold to or operated by an end user, but which may be associated with or operated for the benefit of a user (e.g., a smart meter).

[0186] UE 2200 includes processing circuitry 2202, which is operatively coupled via bus 2204 to input / output interface 2206, power supply 2208, memory 2210, communication interface 2212, and / or any other component or any combination thereof. Some UEs may use... Figure 18 All components or subsets thereof are shown. The degree of integration between components may vary from UE to UE. In addition, some UEs may contain multiple instances of a component, such as multiple processors, memory, transceivers, transmitters, receivers, etc.

[0187] Processing circuitry 2202 is configured to process instructions and data and can be configured to implement any sequential state machine that operatively executes instructions stored in memory 2210 as a machine-readable computer program. Processing circuitry 2202 can be implemented as one or more hardware-implemented state machines (e.g., in discrete logic, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), etc.); programmable logic with suitable firmware; one or more stored computer programs, a general-purpose processor (e.g., a microprocessor or digital signal processor (DSP)) with suitable software; or any combination thereof. For example, processing circuitry 2202 may include multiple central processing units (CPUs).

[0188] In this example, the input / output interface 2206 can be configured to provide one or more interfaces to input devices, output devices, or one or more input and / or output devices. Examples of output devices include speakers, sound cards, video cards, displays, monitors, printers, actuators, transmitters, smart cards, other output devices, or any combination thereof. Input devices allow users to capture information into the UE 2200. Examples of input devices include touch-sensitive or presence-sensitive displays, cameras (e.g., digital cameras, digital camcorders, webcams, etc.), microphones, sensors, mice, trackballs, arrow keys, touchpads, scroll wheels, smart cards, etc. Presence-sensitive displays may include capacitive or resistive touch sensors to sense input from the user. Sensors may be accelerometers, gyroscopes, tilt sensors, force sensors, magnetometers, optical sensors, proximity sensors, biosensors, etc., or any combination thereof. Output devices can use the same type of interface port as input devices. For example, a Universal Serial Bus (USB) port can be used to provide both input and output devices.

[0189] In some embodiments, power supply 2208 is configured as a battery or battery pack. Other types of power sources can be used, such as external power sources (e.g., power outlets), photovoltaic devices, or batteries. Power supply 2208 may also include power circuitry for delivering power from power supply 2208 itself and / or external power sources to various components of UE 2200 via input circuitry or interfaces (such as power cables). For example, delivering power can be used to charge power supply 2208. Power circuitry can perform any formatting, conversion, or other modifications on the power from power supply 2208 to make the power suitable for the various components of UE 2200. Memory 2210 may be or be configured to include memory such as random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), disk, optical disk, hard disk, removable disk enclosure, flash drive, etc. In one example, memory 2210 includes one or more applications 2214, such as an operating system, web browser application, widget, gadget engine, or other application, and corresponding data 2216. The memory 2210 can store any one or a combination of various operating systems for use by the UE2200.

[0190] Memory 2210 can be configured to include multiple physical drive units, such as a Redundant Array of Independent Disks (RAID), flash memory, a USB flash drive, an external hard drive, a thumb drive, a pen drive, a key drive, a high-density digital versatile optical disc (HD-DVD) drive, an internal hard drive, a Blu-ray disc drive, a holographic digital data storage (HDDS) disc drive, an external micro dual in-line memory module (DIMM), synchronous dynamic random access memory (SDRAM), an external micro DIMM (SDRAM), smart card memory (e.g., a tamper-proof module in the form of a universal integrated circuit card (UICC) containing one or more user identity modules (SIMs), such as USIM and / or ISIM), other memory, or any combination thereof. The UICC can be an embedded UICC (eUICC), an integrated UICC (iUICC), or a removable UICC commonly referred to as a "SIM card." Memory 2210 allows UE 2200 to access instructions, applications, etc., stored on temporary or non-temporary storage media to download or upload data. Articles of manufacture, such as those utilizing communication systems, may be tangibly embodied in or contained in memory 2210, which may be or contain a device-readable storage medium.

[0191] Processing circuitry 2202 can be configured to communicate with an access network or other network using communication interface 2212. Communication interface 2212 may include one or more communication subsystems and may include or be communicatively coupled to antenna 2222. Communication interface 2212 may include one or more transceivers for communication, such as through one or more remote transceivers communicating with another device capable of wireless communication (e.g., another UE or a network node in the access network). Each transceiver may include a transmitter 2218 and / or a receiver 2220 suitable for providing network communication (e.g., optical communication, electrical communication, frequency allocation, etc.). Furthermore, transmitter 2218 and receiver 2220 may be coupled to one or more antennas (e.g., antenna 2222) and may share circuitry, software, or firmware, or alternatively be implemented separately.

[0192] In the illustrated embodiment, the communication functions of the communication interface 2212 may include cellular communication, Wi-Fi communication, LPWAN communication, data communication, voice communication, multimedia communication, short-range communication (e.g., Bluetooth), near-field communication, location-based communication (such as using a Global Positioning System (GPS) to determine location), other similar communication functions, or any combination thereof. Communication may be implemented according to one or more communication protocols and / or standards, such as IEEE 802.11, Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), GSM, LTE, New Radio (NR), UMTS, WiMax, Ethernet, Transmission Control Protocol / Internet Protocol (TCP / IP), Synchronous Fiber Network (SONET), Asynchronous Transfer Mode (ATM), QUIC, Hypertext Transfer Protocol (HTTP), etc.

[0193] Regardless of the sensor type, the UE can wirelessly transmit the data captured by its sensors to the network node via its communication interface 2212. Data captured by the UE's sensors can also be wirelessly transmitted to the network node via another UE. The output can be periodic (e.g., every 15 minutes if it reports sensed temperature), random (e.g., to balance the reporting load from multiple sensors), triggered by an event (e.g., sending an alarm when moisture is detected), responded to a request (e.g., a user-initiated request), or a continuous stream (e.g., real-time video feed of a patient).

[0194] As another example, the UE includes an actuator, motor, or switch. This is associated with a communication interface configured to receive wireless input from a network node via a wireless connection. The state of the actuator, motor, or switch may change in response to the received wireless input. For example, the UE may include a motor that adjusts the control surfaces or rotors of a drone in flight based on received input, or it may include a robotic arm that performs a medical procedure based on received input.

[0195] When the UE is an Internet of Things (IoT) device, it can be a device used in one or more application areas, including but not limited to urban wearable technology, extended industrial applications, and healthcare. Non-limiting examples of such IoT devices include: connected refrigerators or freezers, televisions, connected lighting devices, electricity meters, robotic vacuum cleaners, voice-controlled smart speakers, home security cameras, motion detectors, thermostats, smoke detectors, door and window sensors, flood / humidity sensors, power door locks, connected doorbells, air conditioning systems (e.g., heat pumps), autonomous vehicles, surveillance systems, weather monitoring devices, vehicle parking monitoring devices, electric vehicle charging stations, smartwatches, fitness trackers, head-mounted displays for augmented reality (AR) or virtual reality (VR), wearable devices for enhancing tactile or sensory experiences, sprinklers, animal or object tracking devices, sensors for monitoring plants or animals, industrial robots, unmanned aerial vehicles (UAVs), and any type of medical device (e.g., heart rate monitors or remote-controlled surgical robots). In addition to... Figure 18 In addition to the other components described in relation to UE 2200, the UE in the form of an IoT device also includes circuitry and / or software depending on the intended application of the IoT device.

[0196] As another concrete example, in IoT scenarios, a UE can represent a machine or other device that performs monitoring and / or measurement and transmits the results of such monitoring and / or measurement to another UE and / or network node. In this case, the UE can be an M2M device, which can be referred to as an MTC device in the 3GPP context. As a specific example, the UE can implement the 3GPP NB-IoT standard. In other scenarios, a UE can represent a vehicle, such as a car, bus, truck, ship, and aircraft, or other device capable of monitoring and / or reporting its operational status or other functions related to its operation.

[0197] In practice, any number of UEs can be used together for a single use case. For example, the first UE can be a drone or integrated into a drone, providing the drone's speed information (acquired via a speed sensor) to a second UE, which acts as a remote controller for operating the drone. When the user makes changes via the remote controller, the first UE can adjust the drone's throttle (e.g., by controlling actuators) to increase or decrease the drone's speed. The first UE and / or the second UE can also include multiple of the aforementioned functions. For example, a UE can include sensors and actuators and handle data communication between the speed sensors and actuators.

[0198] Figure 19 A network node 3300 according to some embodiments is illustrated. As used herein, "network node" means a device capable of, configured, arranged, and / or operated for communicating directly or indirectly with a UE and / or other network nodes or devices in a telecommunications network. Examples of network nodes include, but are not limited to, access points (APs) (e.g., wireless access points), base stations (BSs) (e.g., wireless base stations, Node Bs, evolved Node Bs (eNBs), and NR Node Bs (gNBs).

[0199] Base stations can be classified according to the coverage they provide (or, in other words, their transmit power level); therefore, based on the coverage provided, a base station can be called a femtobase, picobase, microbase, or macrobase. A base station can be a relay node or a relay donor node controlling a relay. Network nodes can also include one or more (or all) parts of a distributed radio base station, such as a centralized digital unit and / or a remote radio unit (RRU), sometimes called a remote radio headend (RRH). Such a remote radio unit may or may not be integrated with an antenna as part of an antenna-integrated radio system. The various parts of a distributed radio base station can also be referred to as nodes in a distributed antenna system (DAS).

[0200] Other examples of network nodes include multi-TRP 5G access nodes, multi-standard radio (MSR) devices (such as MSR BS), network controllers such as Radio Network Controllers (RNC) or Base Station Controllers (BSC), Base Transceivers (BTS), transport points, transport nodes, Multi-Cell / Multicast Coordination Entities (MCE), Operation and Maintenance (O&M) nodes, Operations Support System (OSS) nodes, Self-Organizing Network (SON) nodes, and location nodes (such as Evolved Services Mobile Location Centers (E-SMLC) and / or Minimized Drive Tests (MDT)).

[0201] Network node 3300 includes processing circuitry 3302, memory 3304, communication interface 3306, and power supply 3308. Network node 3300 may consist of multiple physically independent components (e.g., NodeB components and RNC components, or BTS components and BSC components, etc.), each of which may have its own components. In some scenarios where network node 3300 includes multiple independent components (e.g., BTS and BSC components), one or more independent components may be shared among multiple network nodes. For example, a single RNC can control multiple NodeBs. In this case, each unique NodeB and RNC pair may be considered a single independent network node in some circumstances. In some embodiments, network node 1300 may be configured to support multiple Radio Access Technologies (RATs). In such embodiments, some components may be duplicated (e.g., independent memory 3304 for different RATs), and some components may be reused (e.g., the same antenna 3310 may be shared by different RATs). Network node 3300 may also include multiple sets of graphical components integrated into network node 1300 to support different wireless technologies, such as...

[0202] GSM, WCDMA, LTE, NR, WiFi, Zigbee, Z-wave, LoRaWAN, RFID, or Bluetooth wireless technologies. These wireless technologies can be integrated into the same or different chips or chipsets and other components within the network node 1300.

[0203] Processing circuitry 3302 may include one or more of a combination of a microprocessor, controller, microcontroller, central processing unit, digital signal processor, application-specific integrated circuit, field-programmable gate array, or any other suitable computing device, resource, or operable hardware, software, and / or coding logic, used alone or in conjunction with other network node 3300 components (e.g., memory 3304) to provide the functionality of network node 3300.

[0204] In some embodiments, the processing circuitry 3302 includes a system-on-a-chip (SOC). In some embodiments, the processing circuitry 3302 includes one or more radio frequency (RF) transceiver circuits 3312 and baseband processing circuits 3314. In some embodiments, the RF transceiver circuits 3312 and the baseband processing circuits 3314 may be located on separate chips (or chipsets), circuit boards, or units (e.g., radio units and digital units). In alternative embodiments, some or all of the RF transceiver circuits 3312 and the baseband processing circuits 3314 may be located on the same chip or chipset, circuit board, or unit.

[0205] Memory 3304 may include any form of volatile or non-volatile computer-readable memory, including but not limited to persistent memory, solid-state memory, remotely mounted memory, magnetic media, optical media, random access memory (RAM), read-only memory (ROM), mass storage media (e.g., hard disk), removable storage media (e.g., flash drive, optical disc (CD) or digital video disc (DVD)) and / or any other volatile or non-volatile, non-transitory device-readable and / or computer-executable memory device, for storing information, data, and / or instructions usable by processing circuitry 3302. Memory 3304 may store any suitable instructions, data, or information, including computer programs, software, and application programs, including one or more of logic, rules, codes, tables, and / or other instructions executable by processing circuitry 3302 and utilized by network node 3300. Memory 3304 may be used to store any calculations performed by processing circuitry 3302 and / or any data received through communication interface 3306. In some embodiments, processing circuitry 3302 and memory 3304 are integrated together.

[0206] Communication interface 3306 is used for wired or wireless communication of signaling and / or data between network nodes, access networks, and / or UEs. As shown, communication interface 3306 includes one or more ports / terminals 3316 for sending and receiving data to and from the network, for example, via a wired connection. Communication interface 3306 also includes radio front-end circuitry 3318, which may be coupled to antenna 3310, or in some embodiments as part of antenna 3310. Radio front-end circuitry 3318 includes filter 3320 and amplifier 3322. Radio front-end circuitry 3318 may be connected to antenna 3310 and processing circuitry 3302. Radio front-end circuitry 3318 may be configured to modulate the signal for communication between antenna 3310 and processing circuitry 3302. Radio front-end circuitry 3318 may receive digital data to be transmitted to other network nodes or UEs via a wireless connection. Radio front-end circuitry 3318 may use a combination of filter 3320 and / or amplifier 3322 to convert the digital data into a radio signal with appropriate channel and bandwidth parameters. This radio signal may then be transmitted through antenna 3310. Similarly, when receiving data, antenna 3310 can collect radio signals, which are then converted into digital data by radio front-end circuitry 3318. The digital data can then be passed to processing circuitry 3302. In other embodiments, the communication interface may include different components and / or different combinations of components.

[0207] In some alternative embodiments, network node 3300 does not include a separate radio front-end circuitry 3318; instead, processing circuitry 3302 includes the radio front-end circuitry and is connected to antenna 3310. Similarly, in some embodiments, all or part of the radio frequency transceiver circuitry 3312 is part of communication interface 3306. In other embodiments, communication interface 3306 includes one or more ports or terminals 3316, radio front-end circuitry 3318, and radio frequency transceiver circuitry 3312 as part of a radio unit (not shown), and communication interface 3306 communicates with baseband processing circuitry 3314, which is part of a digital unit (not shown).

[0208] Antenna 3310 may include one or more antennas or an antenna array for transmitting and / or receiving wireless signals. Antenna 3310 may be coupled to radio front-end circuitry 3318 and may be any type of antenna capable of wirelessly transmitting and receiving data and / or signals. In some embodiments, antenna 3310 is decoupled from network node 3300 and may be connected to network node 3300 via an interface or port.

[0209] Antenna 3310, communication interface 3306, and / or processing circuitry 3302 may be configured to perform any receive operation and / or certain acquire operation as described herein by a network node. Any information, data, and / or signal can be received from the UE, another network node, and / or any other network device. Similarly, antenna 3310, communication interface 3306, and / or processing circuitry 3302 may be configured to perform any transmission operation as described herein by a network node. Any information, data, and / or signal can be transmitted to the UE, another network node, and / or any other network device.

[0210] Power supply 3308 supplies power to the various components of network node 3300 in a form suitable for each component (e.g., with appropriate voltage and current). Power supply 3308 may also include or be coupled to power management circuitry to provide power to the components of network node 3300 to perform the functions described herein. For example, network node 3300 may be connected to an external power source (e.g., mains, power outlet) via input circuitry or an interface (e.g., cable), thereby allowing the external power source to power the power supply circuitry of power supply 3308. As another example, power supply 3308 may include a power source in the form of a battery or battery pack connected to or integrated into the power supply circuitry. The battery can provide backup power if the external power source fails.

[0211] Embodiments of network node 3300 may include, except Figure 19 Additional components beyond those shown may be used to provide certain aspects of the network node's functionality, including any of the functions described herein and / or any functionality required to support the topics described herein. For example, network node 3300 may include a user interface device to allow information to be input into and output from network node 3300. This allows users to perform diagnostic, maintenance, repair, and other management functions on network node 3300.

[0212] Figure 20 This is a block diagram of host 4400 according to the various aspects described herein, host 4400 may be Figure 17 An embodiment of host 2116. As used herein, host 4400 can be or includes various combinations of hardware and / or software, including processing resources in a standalone server, blade server, cloud-implemented server, distributed server, virtual machine, container, or server cluster. Host 4400 can provide one or more services to one or more UEs.

[0213] Host 4400 includes processing circuitry 4402 operably coupled via bus 4404 to input / output interface 4406, network interface 4408, power supply 4410, and memory 4412. Other embodiments may include additional components. The features of these components may differ from those described in the previous figures (e.g., Figure 18 and 19 The features described in the devices in the ) are basically similar, so the description is generally applicable to the corresponding components of the host 4400.

[0214] Memory 4412 may include one or more computer programs, including one or more host applications 4414 and data 4416, which may include user data, such as data generated by the UE for the host 4400 or data generated by the host 4400 for the UE. Embodiments of the host 4400 may use all or only a subset of the components shown. The host application 4414 may be implemented based on a container architecture and may provide support for video codecs (e.g., Multi-Function Video Coding (VVC), High Efficiency Video Coding (HEVC), Advanced Video Coding (AVC), MPEG, VP9) and audio codecs (e.g., FLAC, Advanced Audio Coding (AAC), MPEG, G.711), including transcoding for a variety of different categories, types, or UE implementations (e.g., mobile phones, desktop computers, wearable display systems, head-up display systems). The host application 4414 may also provide user authentication and permission checks and may periodically report health status, routing, and content availability to a central node (e.g., a device within the core network or at the edge). Therefore, host 4400 can select and / or instruct different hosts to provide over-the-top services for the UE. Host application 4414 can support various protocols, such as HTTP Real-Time Streaming (HLS), Real-Time Messaging Protocol (RTMP), Real-Time Streaming Protocol (RTSP), HTTP-based Dynamic Adaptive Streaming (MPEG-DASH), etc.

[0215] Figure 21 This is a block diagram illustrating a virtualized environment 5500 in which some embodiments of functionality can be virtualized. In this context, virtualization means creating a virtual version of a device or apparatus, which may include a virtualized hardware platform, storage devices, and network resources. As used herein, virtualization can be applied to any device or component thereof described herein and relates to an implementation in which at least a portion of the functionality is implemented as one or more virtual components. Some or all of the functionality described herein can be implemented as virtual components executed by one or more virtual machines (VMs) implemented in one or more virtual environments 5500, which are hosted by one or more hardware nodes (e.g., hardware computing devices operating as network nodes, UEs, core network nodes, or hosts). Furthermore, in embodiments where the virtual node does not require a radio connection (e.g., a core network node or host), the node can be fully virtualized.

[0216] Application 5502 (also referred to as software instance, virtual device, network function, virtual node, virtual network function, etc.) runs in virtualization environment 5500 to implement some of the features, functions and / or benefits of some embodiments disclosed herein.

[0217] Hardware 5504 includes processing circuitry, memory storing software and / or instructions executable by the hardware processing circuitry, and / or other hardware devices described herein, such as network interfaces, input / output interfaces, etc. The processing circuitry can execute software to instantiate one or more virtualization layers 5506 (also referred to as hypervisors or virtual machine monitors (VMMs)), providing VMs 5508a and 5508b (one or more of which may be collectively referred to as VM 5508), and / or performing any functionality, features, and / or benefits related to some embodiments described herein. Virtualization layer 5506 can present a virtual operating platform that appears as networked hardware to VM 5508.

[0218] VM 5508 includes virtual processing, virtual memory, virtual networking or interfaces, and virtual storage, and can be run by a corresponding virtualization layer 5506. Different embodiments of virtual appliance 5502 instances can be implemented on one or more VM 5508s, and can be implemented in different ways. In some cases, hardware virtualization is referred to as Network Functions Virtualization (NFV). NFV can be used to consolidate various types of network devices onto industry-standard high-capacity server hardware, physical switches, and physical storage, which can reside on equipment in data centers and customer premises.

[0219] In an NFV environment, a VM 5508 can be a software implementation of a physical machine, and its running programs behave as if they were executing on a physical, non-virtualized machine. Each VM 5508 and the program executing it...

[0220] The hardware 5504 portion of a VM (whether it's hardware dedicated to that VM or hardware shared by that VM with other VMs) constitutes a separate virtual network element. Within the NFV environment, the virtual network function is responsible for handling specific network functions running on one or more VMs 5508 on the hardware 5504, and corresponds to the application 5502.

[0221] Hardware 5504 can be implemented in a standalone network node with general-purpose or special-purpose components. Hardware 5504 may implement certain functions through virtualization. Alternatively, hardware 5504 may be part of a larger hardware cluster (e.g., in a data center or CPE) where many hardware nodes work collaboratively and are managed by management and orchestration 5510, which oversees the lifecycle management of application 5502. In some embodiments, hardware 5504 is coupled to one or more radio units, each including one or more transmitters and one or more receivers, which may be coupled to one or more antennas. The radio units can communicate directly with other hardware nodes through one or more suitable network interfaces and can be used in conjunction with virtual components to provide virtual nodes with radio capabilities, such as wireless access nodes or base stations. In some embodiments, a control system 5512 may be used to provide certain signaling, which can alternatively be used for communication between the hardware nodes and radio units.

[0222] Figure 22 A communication diagram is shown illustrating how host 6602 communicates with UE 6606 via a partial wireless connection through network node 6604, according to some embodiments. Reference will now be made to... Figure 22 To describe the UEs (e.g., those discussed in the preceding paragraphs according to various embodiments) Figure 17 UE 2112a and / or Figure 18 UE 2200), network nodes (e.g. Figure 17 Network node 2110a and / or Figure 19 Network node 3300) and host (e.g. Figure 17 Host 2116 and / or Figure 20 Example implementation of host 4400.

[0223] Similar to host 4400, embodiments of host 6602 include hardware such as a communication interface, processing circuitry, and memory. Host 6602 also includes software stored in or accessible by host 6602 and executable by the processing circuitry. This software includes a host application that can be used to provide services to remote users, such as UE 6606 connected via an OTT connection 6650 extending between UE 6606 and host 6602. When providing services to remote users, the host application can provide user data transmitted using the OTT connection 6650.

[0224] Network node 6604 includes hardware that enables it to communicate with host 6602 and UE 6606. Connection 6660 can be a direct connection or via a core network (such as...). Figure 17 The UE 6606 connects to the core network 2106 and / or one or more other intermediate networks (e.g., one or more public networks, private networks, or hosted networks). For example, an intermediate network could be a backbone network or the Internet. The UE 6606 includes hardware and software stored in or accessible by the UE 6606 and executable by the UE's processing circuitry. The software includes client applications, such as web browsers or operator-specific "applications," which can be used to provide services to human or non-human users via the UE 6606 with the support of the host 6602. In the host 6602, the executing host application can communicate with the executing client application via an OTT connection 6650 terminated between the UE 6606 and the host 6602. When providing services to users, the UE's client application can receive request data from the host application of the host and provide user data in response to the request data. The OTT connection 6650 can transmit request data and user data. The UE's client application can interact with users to generate user data and provide it to the host application via the OTT connection 6650.

[0225] OTT connection 6650 can be extended via connection 6660 between host 6602 and network node 6604 and wireless connection 6670 between network node 6604 and UE 6606 to provide connectivity between host 6602 and UE 6606. Connection 6660 and wireless connection 6670 (through which OTT connection 6650 can be provided) have been abstractly drawn to illustrate communication between host 6602 and UE 6606 via network node 6604, without explicitly mentioning any intermediate devices or the precise routing of messages via these devices.

[0226] As an example of data transmission via OTT connection 6650, in step 6608, host 6602 provides user data, which can be performed by executing a host application. In some embodiments, the user data is associated with a specific human user interacting with UE 6606. In other embodiments, the user data is associated with UE 6606, which shares data with host 6602 without explicit human interaction. In step 6610, host 6602 initiates a transmission carrying user data to UE 6606. Host 6602 may initiate the transmission in response to a request sent by UE 6606. This request may be caused by human interaction with UE 6606 or by the operation of a client application running on UE 6606. According to the teachings of the embodiments described in this disclosure, this transmission can be performed via network node 6604. Therefore, in step 6612, according to the teachings of the embodiments described in this disclosure, network node 6604 transmits the user data carried in the transmission initiated by host 6602 to UE 6606. In step 6614, UE 6606 receives user data carried in the transmission, which can be executed by a client application running on UE 6606, which is associated with a host application running on host 6602.

[0227] In some examples, UE 6606 executes a client application that provides user data to host 6602. The user data may be provided as a response or reaction to data received from host 6602. Therefore, in step 6616, UE 6606 can provide user data, which can be performed by executing the client application. When providing user data, the client application may also consider user input received from a user via the input / output interface of UE 6606. Regardless of the specific manner in which the user data is provided, UE 6606 initiates a transmission of user data to host 6602 via network node 6604 in step 6618. In step 6620, in accordance with the teachings of the embodiments described in this disclosure, network node 6604 receives user data from UE 6606 and initiates a transmission of the received user data to host 6602. In step 6622, host 6602 receives the user data carried in the transmission initiated by UE 6606.

[0228] One or more embodiments improve the performance of OTT services provided to UE 6606 using OTT connection 6650 (where wireless connection 6670 constitutes the final segment). More specifically, the teachings of these embodiments can improve data rates, latency, and / or power consumption, thereby providing benefits such as reduced user wait times, relaxed file size limits, improved content resolution, improved response times, and / or extended battery life.

[0229] In the example scenario, factory status information can be collected and analyzed by host 6602. As another example, host 6602 can process audio and video data obtained from the UE to create maps. As another example, host 6602 can collect and analyze real-time data to assist in controlling traffic congestion (e.g., controlling traffic lights). As another example, host 6602 can store surveillance video uploaded by the UE. As another example, host 6602 can store or control access to media content such as video, audio, VR, or AR broadcast, multicast, or unicast to the UE. As other examples, host 6602 can be used for energy pricing, remote control of non-time-critical power loads to balance generation demand, location services, presentation services (such as compiling charts based on data collected from remote devices), or any other function that collects, retrieves, stores, analyzes, and / or transmits data.

[0230] In some examples, measurement procedures may be provided for monitoring data rates, latency, and other factors that improve upon one or more embodiments. Additionally, optional network functions may be provided for reconfiguring the OTT connection 6650 between host 6602 and UE 6606 in response to changes in measurement results. The measurement process and / or the network functions for reconfiguring the OTT connection may be implemented in the software and hardware of host 6602 and / or UE 6606. In some embodiments, sensors (not shown) may be deployed in or associated with other devices traversed by the OTT connection 6650; the sensors may participate in the measurement process by providing values ​​of the aforementioned monitored quantities or by providing values ​​of other physical quantities that the software can calculate or estimate based on them. Reconfiguration of the OTT connection 6650 may include message formats, retransmission settings, preferred routing, etc.; reconfiguration does not require direct changes to the operation of network node 6604. Such processes and functions are known and practiced in the art. In some embodiments, the measurement may involve proprietary UE signaling, which helps host 6602 measure throughput, propagation time, latency, etc. Measurements can be performed using software that uses an OTT connection to transmit messages via the 6650, particularly empty or "virtual" messages, while simultaneously monitoring propagation time, errors, and other factors.

[0231] While the computing devices described herein (e.g., UE, network node, host) may include the illustrated combinations of hardware components, other embodiments may include computing devices with different combinations of components. It should be understood that these computing devices may include any suitable hardware and / or software combination required to perform the tasks, features, functions, and methods disclosed herein. The determination, computation, acquisition, or similar operations described herein may be performed by processing circuitry that may, for example, convert acquired information into other information, compare the acquired or converted information with information stored in a network node, and / or perform one or more operations based on the acquired or converted information, and make a determination based on the result of said processing. Furthermore, although components are depicted as a single box within a larger box, or nested within multiple boxes, in practice, a computing device may contain multiple different physical components constituting a single illustrated component, and functionality may be partitioned between individual components. For example, a communication interface may be configured to include any of the components described herein, and / or the functionality of the components may be partitioned between processing circuitry and the communication interface. In another example, non-computationally intensive functions of any such component may be implemented in software or firmware, while computationally intensive functions may be implemented in hardware.

[0232] In some embodiments, some or all of the functionality described herein may be provided by processing circuitry that executes instructions stored in memory, which may be a computer program product in the form of a non-transitory computer-readable storage medium. In alternative embodiments, some or all of the functionality may be provided by processing circuitry without executing instructions stored on a separate or discrete device-readable storage medium (such as in a hard-wired manner). In any of these particular embodiments, the processing circuitry may be configured to perform the described functionality regardless of whether instructions stored on a non-transitory computer-readable storage medium are executed. The benefits provided by such functionality are not limited to the processing circuitry itself or other components of the computing device, but are shared by the entire computing device and / or the end user and the entire wireless network.

[0233] Abbreviations and Definitions

[0234] To aid in understanding the scope and content of this specification and the appended claims, some terms are defined directly below. 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.

[0235] The terms “approximately,” “about,” and “substantially” used herein refer to quantities or conditions that are close to a specific specification but still perform the desired function or achieve the desired result. For example, the terms “approximately,” “about,” and “substantially” may refer to quantities or conditions that deviate from a specific specification by less than 10%, less than 5%, less than 1%, less than 0.1%, or less than 0.01%.

[0236] Various aspects of this disclosure, including apparatus, systems, and methods, may be described with reference to one or more embodiments or implementations that are exemplary in nature. The term "exemplary" as used herein means "serving as an example, instance, or illustration" and should not necessarily be construed as superior to or better than other embodiments disclosed herein. Furthermore, references to "implementation" of this disclosure or embodiments include specific references to one or more embodiments thereof, and are intended to provide illustrative examples without limiting the scope of this disclosure, which is defined by the appended claims rather than this specification.

[0237] As used in this specification, singular words include their plural counterparts, and plural words include their singular counterparts, unless otherwise expressly stated or implied. Therefore, it should be noted that, as used in this specification and the appended claims, unless the context clearly specifies otherwise, the singular forms "a," "an," and "the" all include plural referents. For example, unless otherwise expressly or implicitly understood or stated, a reference to a singular referent (e.g., "a micro-part") includes one, two, or more referents. Similarly, unless the content and / or context clearly specify otherwise, a reference to multiple referents should be interpreted as including a single referent and / or multiple referents. For example, a reference to a plural referent (e.g., "micro-part") does not necessarily require multiple such referents. Rather, it should be understood that, regardless of the inferred number of referents, unless otherwise stated, this document contemplates including one or more referents.

[0238] References to "an embodiment," "one example," "exemplary embodiment," etc., in this specification indicate that the embodiment may include specific features, structures, or characteristics, but not every embodiment includes that specific feature, structure, or characteristic. Furthermore, these expressions do not necessarily refer to the same embodiment. Moreover, when a specific feature, structure, or characteristic is described in connection with an embodiment, it should be understood that the knowledge of those skilled in the art enables the application of such features, structures, or characteristics to other embodiments, whether explicitly described or not.

[0239] It should be understood that although the terms "first" and "second," etc., may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used only to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any combination of one or more of the associated listed terms.

[0240] It should also be understood that the terms "comprising," "including," "having," "having," "containing," and / or "including" as used herein specify the presence of the stated features, elements, and / or components, but do not exclude the presence or addition of one or more other features, elements, components, and / or combinations thereof.

[0241] in conclusion

[0242] This disclosure includes any novel features or combinations of features expressly disclosed herein, whether explicitly disclosed or generalized. Various modifications and alterations to the foregoing exemplary embodiments of this disclosure will be apparent to those skilled in the art in conjunction with the foregoing description. However, any and all modifications should still fall within the scope of the non-limiting and exemplary embodiments of this disclosure.

[0243] It should be understood that, for any given component or embodiment described herein, any possible candidates or alternatives listed for that component may generally be used alone or in combination with each other, unless otherwise expressly or implied. Furthermore, it should be understood that, unless otherwise expressly or implied, any list of such candidates or alternatives is illustrative only and not restrictive.

[0244] Furthermore, unless otherwise stated, the figures used to represent quantities, components, distances, or other measurements in this specification and claims should be understood to be modified by the word "about," which is defined consistent with this document. Therefore, unless otherwise stated, the numerical parameters listed in this specification and the appended claims are approximate values ​​and may vary depending on the desired characteristics to be obtained from the subject matter described herein. At least, without attempting to limit the application of the doctrine of equivalents to the scope of the claims, each numerical parameter should be interpreted based on at least the number of significant figures reported and the application of conventional rounding techniques. While the numerical ranges and parameters described within the broad scope of the subject matter herein are approximate, the values ​​listed in the specific examples have been reported as precisely as possible. However, any numerical value inherently contains some error, which necessarily stems from the standard deviation of its respective test measurement.

[0245] Any headings and subheadings used herein are for organizational purposes only and are not intended to limit the scope of the specification or claims. The terms and expressions used herein are descriptive terms only and not restrictive terms, and the use of such terms and expressions is not intended to exclude any equivalents of the shown and described features or portions thereof, but it should be recognized that various modifications can be made within the scope of this disclosure. Therefore, although this disclosure is partly specific to certain embodiments, those skilled in the art can adopt optional features, modifications, and variations of the concepts disclosed herein, and such modifications and variations are considered to be within the scope of this specification.

[0246] It should also be understood that, according to certain embodiments of this disclosure, systems, apparatuses, products, kits, methods, and / or processes may include, incorporate, or otherwise include features or characteristics (e.g., components, members, elements, parts, and / or portions) described in other embodiments disclosed and / or described herein. Therefore, various features of certain embodiments may be compatible with, combined with, included in, and / or incorporated into other embodiments of this disclosure. Consequently, the disclosure of certain features with respect to a particular embodiment of this disclosure should not be construed as limiting the application or inclusion of said features to that particular embodiment. Rather, it should be understood that other embodiments may also include said features, members, elements, parts, and / or portions without departing from the scope of this disclosure.

[0247] Furthermore, unless a feature is described as requiring another feature in combination with it, any feature herein may be combined with any other feature of the same or different embodiments disclosed herein. Moreover, to avoid obscuring aspects of the exemplary embodiments, various well-known aspects of exemplary systems, methods, apparatuses, etc., are not specifically described in detail herein. However, such aspects are also contemplated herein.

[0248] It will be apparent to those skilled in the art that, in addition to the methods, apparatuses, apparatus elements, materials, procedures, and techniques specifically described herein, other methods, apparatuses, apparatus elements, materials, procedures, and techniques can also be applied to the practice of the embodiments broadly disclosed herein without requiring excessive experimentation. All functional equivalents known in the art of the methods, apparatuses, apparatus elements, materials, procedures, and techniques specifically described herein are intended to be covered within this disclosure.

[0249] When a group of materials, compositions, components, or compounds is disclosed herein, it should be understood that all individual members of that group and all its subgroups are disclosed separately. When the Markush group or other groupings are used herein, all individual members of that group and all possible combinations and subcombinations of that group are intended to be included separately in this disclosure.

[0250] The above embodiments are merely examples. Those skilled in the art can make changes, modifications, and variations to specific embodiments without departing from the scope of the specification, which is defined only by the appended claims.

[0251] References

[0252] 1. Joint Antenna Selection and Hybrid Beamforming Design using Unquantized and Quantized Deep Learning Networks; Ahmet M. Elbir and Kumar Vijay Mishra; IEEE Transactions on Wireless Communications 2020.

[0253] 2.Unsupervised Deep Learning for Massive MIMO Hybrid Beamforming; Hamed Hojatian, Jérémy Nadal, Jean- Frigon and Leduc-Primeau; IEEE Transactions on Wireless Communications 2021.

[0254] 3. Deep Unsupervised Learning for Joint Antenna Selection and HybridBeamforming; Zhiyan Liu, Yuwen Yang, Feifei Gao, Ting Zhou and Hongbing Ma; IEEE Transactions on Communications 2022.

[0255] 4. Flexible Unsupervised Learning for Massive MIMO Subarray HybridBeamforming; Hamed Hojatian, Jeremy Nadal, Jean-Francois Frigon, and Francois Leduc-Primeau; IEEE GLOBECOM 2022.

[0256] 5. PrecoderNet: Hybrid Beamforming for Millimeter Wave System with Deep Reinforcement Learning; Qisheng Wang, Keming Feng, Xiao Li and Shi Jin; IEEE Wireless Communications Letters, 2020.

[0257] 6. Sub-Array Hybrid Precoding for Massive MIMO Systems: A CNN-Based Approach; Kai Chen, Jing Yang, Qiang Li and Xiaho Ge; IEEE Communications Letters, 2021.

[0258] 7. Dynamic Subarray for Hybrid Precoding in Wideband mmWave MIMO Systems; Sungwoo Park, Ahmed Alkhateeb, and Robert W. Heath; IEEE Transactions on Wireless Communications, 2017.< / c> < / c>

Claims

1. A method performed by a base station including multiple antennas for performing beamforming in a massive multiple-input multiple-output (MIMO) system, the method comprising: Measure the power consumption and insertion loss of one or more components that make up a base station; Energy consumption is determined based on the power consumption and insertion loss. Detect the number of user equipment (UE) communicating with the base station; Measure the spectral efficiency of the base station; Energy consumption, number of UEs, and spectral efficiency are compared with one or more outputs of a trained machine learning model, wherein the trained machine learning model is trained using a smart loss function that is at least partially based on one or more data related to energy consumption, spectral efficiency, and number of UEs. as well as Based on the comparison results, one or more beamforming structures are created using multiple antennas.

2. The method according to claim 1, wherein the one or more data includes at least one of the following: channel state information; imperfect channel state information.

3. The method of claim 1 or 2, wherein the training comprises using one or more DNNs (deep neural networks), wherein the one or more DNNs comprise a plurality of convolutional layers, followed by a plurality of fully connected layers, and each layer is followed by batch normalization.

4. The method according to any one of claims 1 to 3, wherein the training comprises unsupervised learning.

5. The method according to any one of claims 1 to 4, wherein the one or more components comprise one or more of the following: one or more combiners; one or more mixers; one or more amplifiers; one or more antennas; one or more low-pass filters; one or more digital-to-analog converters; One or more local oscillators; One or more digital pre-encoders; one or more analog pre-encoders; one or more phase shifters; one or more switches; an RF front end including circuitry between an antenna and a digital-to-analog converter; one or more passive components.

6. The method according to any one of claims 1 to 5, wherein the intelligent loss function comprises the following functions: Loss=-SE+γEC+ζAS in, SE represents spectral efficiency, EC represents energy consumption, AS represents adaptive antenna selection based on desired spectral efficiency, and hyperparameters γ and ζ are used to control the weight of each term to obtain a trade-off between SE and energy efficiency.

7. The method of claim 3, wherein the one or more DNNs comprise a plurality of convolutional layers, followed by a plurality of fully connected layers.

8. The method of claim 7, wherein batch normalization is used after each layer to avoid overfitting.

9. The method of claim 3, 7 or 8, wherein one of the one or more DNNs is used to perform hybrid beamforming.

10. The method according to any one of claims 7 to 9, wherein the output of the two fully connected layers is divided into four output layers, wherein: The first and second of the four output layers are configured to generate the real and imaginary parts of the digital precoder; The third layer is configured to generate an analog precoder; as well as The fourth layer is configured to design binary matrices.

11. The method of claim 10, wherein a differentiable approximation is used to generate the fourth layer.

12. The method of claim 11, wherein the differentiable approximation used is the Gumbel-Sigmoid approximation, wherein the Gumbel-Sigmoid function is based at least in part on the Gumbel-Softmax equation G(Π), applied to each element of matrix Π, and is defined as: Where Ω HB is the output of the DNN, and g and g' are independent samples that follow a Gumbel distribution with zero mean and unit variance.

13. The method according to any one of claims 10 to 12, wherein the third layer is adapted to different phase shifter resolutions.

14. The method of claim 6, wherein the spectral efficiency of the loss function corresponds to maximizing the spectral efficiency by optimizing one or more analog precoders and one or more digital precoders.

15. The method of claim 6, wherein the loss function is for N RF <<N T Hybrid beamforming and having N RF =N T The all-digital precoder is defined.

16. The method of claim 6, wherein the term γEC of the loss function corresponds to reducing the power consumption of the BS.

17. The method of claim 6, wherein the term ζAS of the loss function corresponds to beamforming designed based on the number of active UEs, and guarantees the minimum energy required to achieve the desired average rate.

18. The method of claim 6, wherein parameters γ and ζ provide a trade-off between energy consumption and spectral efficiency.

19. A method performed by a base station including multiple antennas for performing beamforming in a massive multiple-input multiple-output (MIMO) system, the method comprising: Measure the power consumption and insertion loss of one or more components that make up a base station; Energy consumption is determined based on the power consumption and insertion loss. Detect the number of user-configured devices (UEs) communicating with the base station; Measure the spectral efficiency of the base station; The machine learning model is trained using a smart loss function, based at least in part on one or more data related to the energy consumption, spectral efficiency, and number of UEs. as well as Based on a trained machine learning model, one or more beamforming structures are created using multiple antennas.

20. The method of claim 19, wherein the one or more data includes at least one of: channel state information; imperfect channel state information.

21. The method of claim 19 or 20, wherein the training comprises using one or more DNNs (deep neural networks), wherein the one or more DNNs comprise a plurality of convolutional layers, followed by a plurality of fully connected layers, and each layer is followed by batch normalization.

22. The method according to any one of claims 19 to 21, wherein the training comprises unsupervised learning.

23. The method according to any one of claims 19 to 22, wherein the one or more components comprise one or more of the following: one or more combiners; one or more mixers; one or more amplifiers; one or more antennas; one or more low-pass filters; one or more digital-to-analog converters; One or more local oscillators; One or more digital pre-encoders; one or more analog pre-encoders; one or more phase shifters; one or more switches; an RF front end including circuitry between an antenna and a digital-to-analog converter; one or more passive components.

24. The method according to any one of claims 19 to 23, wherein the intelligent loss function comprises the following function: Loss=-SE+γEC+ζAS in, SE represents spectral efficiency, EC represents energy consumption, AS represents adaptive antenna selection based on desired spectral efficiency, and hyperparameters γ and ζ are used to control the weight of each term to obtain a trade-off between SE and energy efficiency.

25. The method of claim 21, wherein the one or more DNNs comprise a plurality of convolutional layers, followed by a plurality of fully connected layers.

26. The method of claim 25, wherein batch normalization is used after each layer to avoid overfitting.

27. The method of claim 21, 25 or 26, wherein one of the one or more DNNs is used to perform hybrid beamforming.

28. The method according to any one of claims 25 to 27, wherein the outputs of the plurality of fully connected layers are divided into four output layers, wherein: The first and second of the four output layers are configured to generate the real and imaginary parts of the digital precoder; The third layer is configured to generate an analog precoder; as well as The fourth layer is configured to design binary matrices.

29. The method of claim 28, wherein a differentiable approximation is used to generate the fourth layer.

30. The method of claim 29, wherein the differentiable approximation used is the Gumbel-Sigmoid approximation, wherein the Gumbel-Sigmoid function is based at least in part on the Gumbel-Softmax equation G(Π), applied to each element of matrix Π, and is defined as: Where Ω HB is the output of the DNN, and g and g' are independent samples that follow a Gumbel distribution with zero mean and unit variance.

31. The method according to any one of claims 28 to 30, wherein the third layer is adapted to different phase shifter resolutions.

32. The method of claim 24, wherein the spectral efficiency of the loss function corresponds to maximizing the spectral efficiency by optimizing one or more analog precoders and one or more digital precoders.

33. A network node for performing hybrid beamforming or all-digital precoding, the network node comprising: The processing circuit is configured to perform any step of any one of claims 1 to 32; A power supply circuit is configured to supply power to the processing circuit.

Images