Device active state detection method and system based on PSCA algorithm

Abstract

The invention discloses an equipment active state detection method and system based on a PSCA algorithm. The method comprises the following steps: receiving pilot signals sent by a plurality of pieces of equipment to be detected; inputting the pilot signal into a preset PSCA algorithm, calculating a first covariance matrix corresponding to the pilot signal in each iteration, determining a corresponding inverse matrix based on the first covariance matrix, and calculating an optimal solution of an active state estimation optimization problem corresponding to each to-be-detected device in parallel according to the inverse matrix, performing iterative calculation on the active state estimation value according to a preset algorithm step length and the optimal solution, and when a preset iteration termination condition is met, outputting a target active state estimation value; and obtaining an active state detection result of each to-be-detected device based on the target active state estimation value. According to the invention, the accuracy and speed of device active state detection in an unlicensed access scene can be improved.

Description

Technical Field

[0001] The present application relates to the field of massive machine communications, and in particular to a device activity status detection method and system based on a PSCA algorithm. Background Art

[0002] Massive machine type communication (mMTC) and ultra-reliable low-latency communication (URLLC) have become the core application scenarios in 5G and post-5G (B5G) wireless communication systems. These application scenarios face huge challenges in transmission speed and power loss during implementation. Currently, unlicensed access technology is mainly used to solve the problems of mMTC and URLLC. However, in the unlicensed access scenario, the base station needs to detect the status of active devices with non-orthogonal pilots that may conflict with each other. The large number of devices and the changing active status increase the complexity of active device status detection.

[0003] At present, the block coordinate descent (BCD) algorithm, projection gradient (PG) algorithm and approximate message passing (AMP) algorithm are mainly used to detect the active status of devices in the unauthorized access scenario, but these methods all have their defects. Among them, the BCD algorithm needs to serially update the active status of all devices in each iteration, so its calculation time will be significantly extended with the increase of the number of devices; although the PG algorithm can update the active status of all devices in parallel in each iteration, which improves the calculation speed to a certain extent, it only uses the first-order information of the objective function for calculation, so the convergence speed is slow; and the AMP algorithm reduces the calculation complexity by introducing a large number of approximate operations, thereby achieving a faster processing speed, but this approximate processing increases the error rate of device activity detection to a certain extent.

[0004] Application Contents

[0005] The present application provides a device activity status detection method and system based on the PSCA algorithm to improve the accuracy and speed of device activity status detection in unauthorized access scenarios.

[0006] In a first aspect, the present application provides a device activity status detection method based on the PSCA algorithm, comprising:

[0007] Receiving pilot signals sent by several devices to be detected;

[0008] Input the pilot signal into a preset PSCA algorithm, calculate a first covariance matrix corresponding to the pilot signal in each iteration, determine a corresponding inverse matrix based on the first covariance matrix, and calculate the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected in parallel according to the inverse matrix, iteratively calculate the active state estimation value according to a preset algorithm step size and the optimal solution, and output a target active state estimation value when a preset iteration termination condition is met, wherein the active state estimation optimization problem is obtained by approximating the initial active state estimation optimization problem after applying it to each of the devices to be detected;

[0009] Based on the target activity state estimation value, an activity state detection result of each of the devices to be detected is obtained.

[0010] The embodiment of the present application receives pilot signals sent by several devices to be detected, so as to facilitate the subsequent input of a preset PSCA algorithm to accurately and quickly detect the active state of the corresponding device; the pilot signal is input into the preset PSCA algorithm, and the corresponding inverse matrix is ​​determined based on the first covariance matrix, so as to obtain an inverse matrix that can accurately reflect the relationship between the active state of the device and the pilot signal, so as to facilitate the subsequent accurate determination of the target active state estimate; the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix, so as to quickly and accurately calculate the optimal solution corresponding to the active state estimation optimization problem, and then accurately determine the active state of the device; at the same time, the convergence speed and detection accuracy can be balanced by dynamically adjusting the algorithm step size. Therefore, compared with the prior art, the present application can significantly improve the accuracy and speed of device active state detection in the unauthorized access scenario.

[0011] Furthermore, the determining of the corresponding inverse matrix based on the first covariance matrix is ​​specifically:

[0012] The real part and the imaginary part of the complex number are extracted from the first covariance matrix, and an inverse matrix corresponding to the first covariance matrix is ​​calculated based on the real part and the imaginary part.

[0013] In this way, by determining the corresponding inverse matrix based on the first covariance matrix, an inverse matrix that can accurately reflect the relationship between the device activity state and the pilot signal can be obtained, which facilitates the subsequent accurate determination of the target activity state estimation value.

[0014] Furthermore, the calculation formula for determining the corresponding inverse matrix based on the first covariance matrix is ​​specifically:

[0015]

[0016]

[0017]

[0018] In the formula, and Represent the real and imaginary parts of a complex number respectively; is the inverse matrix; is the first covariance matrix; i is an imaginary unit.

[0019] Furthermore, the active state estimation optimization problem includes a first estimation problem, wherein the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix, specifically:

[0020] Determine a corresponding second covariance matrix based on the pilot signal, and determine the first estimation problem for device activity state detection based on the second covariance matrix, the first covariance matrix and the inverse matrix, wherein the first estimation problem is approximated by applying an initial maximum likelihood estimation problem to each of the devices to be detected;

[0021] The first estimation problem is solved to obtain the corresponding optimal solution.

[0022] In this way, by solving the first estimation problem, the optimal solution corresponding to the initial maximum likelihood estimation problem can be quickly and accurately calculated, thereby accurately determining the active state of the device.

[0023] Furthermore, the first estimation problem is solved to obtain a calculation formula for the corresponding optimal solution, which is specifically:

[0024]

[0025]

[0026]

[0027]

[0028] In the formula, is the first estimation problem of the nth device to be detected, α n is the active state of the nth device to be detected, is the estimated value of the active state of the kth iteration; n is the nth device to be detected, is a collection of the devices to be detected; is the first covariance matrix; is the inverse matrix; is the second covariance matrix; g n is the path loss of the channel between the nth device to be detected and the base station; sn A pilot signal sent by the nth device to be detected; is the optimal solution to the first estimation problem.

[0029] Furthermore, the active state estimation optimization problem also includes a second estimation problem, wherein the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix, specifically:

[0030] Determine a corresponding second covariance matrix based on the pilot signal;

[0031] Determine a corresponding probability distribution function based on the activity state distribution of the plurality of devices to be detected, and determine the second estimation problem for device activity state detection based on the second covariance matrix and the probability distribution function, wherein the second estimation problem is obtained by approximating an initial maximum a posteriori estimation problem after applying it to each device to be detected;

[0032] The second estimation problem is solved to obtain the corresponding optimal solution.

[0033] In this way, by solving the second estimation problem, the optimal solution corresponding to the maximum a posteriori estimation problem can be quickly and accurately calculated, thereby accurately determining the active state of the device.

[0034] Furthermore, the second estimation problem is solved to obtain the corresponding calculation formula of the optimal solution, which is specifically:

[0035]

[0036]

[0037]

[0038]

[0039]

[0040] In the formula, is the second estimation problem of the nth device to be detected; α n is the active state of the nth device to be detected, α (k) is the estimated value of the active state of the kth iteration; n is the nth device to be detected, is a collection of the devices to be detected; is the first covariance matrix; is the inverse matrix; is the second covariance matrix; f′ t (α(k) ) is the gradient of the negative logarithm of the probability distribution function of the active state at the kth iteration with respect to the jump state α; c ω Is active state α n The coefficient of correlation, n∈ω, ω∈Ψ, Ψ represents A non-empty subset of n represents the activity probability of the nth device to be detected, where G represents the index of the general model of device activity distribution; I represents the index of the independent model of device activity distribution; g n is the path loss of the channel between the nth device to be detected and the base station; s n A pilot signal sent by the nth device to be detected; is the optimal solution to the second estimation problem.

[0041] Furthermore, the calculation formula for iteratively calculating the activity state estimation value according to the preset algorithm step size and the optimal solution is specifically:

[0042]

[0043] In the formula, and are the estimated values ​​of active states at the k+1th and kth iterations, respectively; is the preset algorithm step size; is the optimal solution of the active state estimation optimization problem; n is the nth device to be detected, is a set of devices to be detected.

[0044] Furthermore, the preset algorithm step size includes a first algorithm step size that satisfies a preset step size condition or a second algorithm step size determined after neural network optimization, specifically:

[0045] The first algorithm step size satisfies In the formula, is the step size of the first algorithm; k is the number of iterations;

[0046] The second algorithm step size is to train the neural network model using the acquired sample data set in a supervised manner until the loss function is minimized, and the model parameters of the neural network model are determined.

[0047] In a second aspect, the present application provides a device activity status detection system based on a PSCA algorithm, comprising: a receiving module, a processing module and a determining module;

[0048] The receiving module is used to receive pilot signals sent by a plurality of devices to be detected;

[0049] The processing module is used to input the pilot signal into a preset PSCA algorithm, calculate a first covariance matrix corresponding to the pilot signal in each iteration, determine a corresponding inverse matrix based on the first covariance matrix, and calculate in parallel the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected according to the inverse matrix, iteratively calculate the active state estimation value according to a preset algorithm step size and the optimal solution, and output a target active state estimation value when a preset iteration termination condition is met, wherein the active state estimation optimization problem is obtained by approximating the initial active state estimation optimization problem after applying it to each of the devices to be detected;

[0050] The determination module is used to obtain the activity state detection result of each of the devices to be detected based on the target activity state estimation value.

[0051] The embodiment of the present application receives pilot signals sent by several devices to be detected, so as to facilitate the subsequent input of a preset PSCA algorithm to accurately and quickly detect the active state of the corresponding device; the pilot signal is input into the preset PSCA algorithm, and the corresponding inverse matrix is ​​determined based on the first covariance matrix, so as to obtain an inverse matrix that can accurately reflect the relationship between the active state of the device and the pilot signal, so as to facilitate the subsequent accurate determination of the target active state estimate; the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix, so as to quickly and accurately calculate the optimal solution corresponding to the active state estimation optimization problem, and then accurately determine the active state of the device; at the same time, the convergence speed and detection accuracy can be balanced by dynamically adjusting the algorithm step size. Therefore, compared with the prior art, the present application can significantly improve the accuracy and speed of device active state detection in the unauthorized access scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flowchart of an embodiment of a device activity status detection method based on the PSCA algorithm provided by the present application;

[0053] Figure 2 It is a structural diagram of the network architecture of the neural network model provided by this application;

[0054] Figure 3 It is a structural diagram of an embodiment of a device activity status detection system based on the PSCA algorithm provided by the present application. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0056] It should be understood that the step numbers used in this article are only for the convenience of description and are not intended to limit the order in which the steps are executed.

[0057] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.

[0058] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0059] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0060] mMTC and URLLC are the core scenarios of 5G / B5G wireless communications, facing the challenges of transmission speed and power loss. Unlicensed access technology can be used to solve these problems, but the base station needs to detect a large number of variable active devices that may conflict under non-orthogonal pilots. The large number of devices and the variable active states increase the complexity of detection. At present, the BCD algorithm, PG algorithm, and AMP algorithm can all be used to detect active states, but each has its own defects. Among them, the BCD algorithm needs to update the device status serially, and the increase in devices will lead to longer calculation time; although the PG algorithm can improve efficiency by updating the active states of all devices in parallel, it only uses first-order information, so the convergence speed is slow; the AMP algorithm reduces complexity through approximate operations, and the processing speed is fast, but the error rate is increased.

[0061] Next, the nouns involved in this application are analyzed:

[0062] Massive Machine Type Communication (mMTC) is a communication technology and service for large-scale Internet of Things (IoT) devices. mMTC is designed to support the communication needs between a large number of IoT devices with low power consumption, low data rate and dormant state most of the time. It is one of the key applications of 5G communication networks and solves the challenges of previous mobile communication technologies in connecting a large number of low-power devices.

[0063] Ultra Reliable & Low Latency Communication (URLLC) is one of the three major application scenarios of 5G communication technology, which is designed to support services that are highly sensitive to latency and stability. Among them, URLLC meets the needs of specific industries and applications by providing extremely high reliability and extremely low latency. This communication technology is crucial for scenarios such as autonomous driving, industrial control, and telemedicine that have extremely high requirements for latency and reliability.

[0064] Grant free (non-orthogonal) multiple access, also known as grant free random access or scheduling free access, is an access method in wireless communication technology. Its core idea is to allow devices to send data without explicit authorization or resource scheduling from the base station, thereby significantly reducing access delay and signaling overhead.

[0065] Parallel Successive Convex Approximation (PSCA) is an optimization algorithm mainly used to solve non-convex optimization problems. The PSCA algorithm combines the ideas of Successive Convex Approximation (SCA) and parallel computing. Its basic idea is to transform the original non-convex optimization problem into a series of easy-to-handle convex subproblems, and approximate the optimal solution of the original problem by parallel computing these convex subproblems.

[0066] The Block Coordinate Descent (BCD) algorithm is an optimization algorithm that is mainly used to solve multivariable optimization problems under unconstrained or multi-constrained conditions. In the BCD algorithm, only one variable is optimized in each iteration, while the other variables remain unchanged. The BCD algorithm divides the variables into several groups (or blocks) and optimizes only one group (block) in each iteration, thereby reducing the amount of calculation.

[0067] The Projection Gradient (PG) algorithm is an iterative algorithm for solving constrained optimization problems. The PG algorithm combines the gradient descent method and the projection operation to solve optimization problems with constraints. In each iteration, the gradient descent method is first used to determine a search direction, which points to the position where the objective function value decreases; then, the points in this search direction are projected into the feasible domain of the constraints to ensure that the solution satisfies the constraints. By continuously iterating this process, the algorithm gradually approaches the optimal solution.

[0068] Approximate Message Passing (AMP) is an iterative algorithm for solving high-dimensional statistical inference problems. The AMP algorithm estimates unknown parameters by iteratively passing information, and has the advantages of high efficiency, simplicity, and scalability. It is mainly used in signal processing, machine learning, computer vision and other fields, especially in large-scale data and sparse data processing. The basic idea of ​​the AMP algorithm is to transform complex inference problems into some simple computational problems, and use some prior knowledge and structural information to estimate unknown parameters.

[0069] Based on this, the embodiments of the present application provide a method and system for detecting the active state of a device based on the PSCA algorithm, which can improve the accuracy and speed of detecting the active state of a device in an unauthorized access scenario.

[0070] The embodiments of the present application provide a device activity status detection method and system based on the PSCA algorithm, which are specifically described through the following embodiments. First, the device activity status detection method based on the PSCA algorithm in the embodiments of the present application is described.

[0071] The device active state detection method based on the PSCA algorithm provided in the embodiment of the present application relates to the field of massive machine communication. The device active state detection method based on the PSCA algorithm provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements a device active state detection method based on the PSCA algorithm, etc., but is not limited to the above forms.

[0072] The present application can also be used in numerous general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0073] Generally speaking, the uplink of a narrowband wireless communication system consists of a base station (BS) with M antennas and an IoT device with N antennas, which is the device to be detected below. The set of antennas and the device to be detected are respectively denoted as and Among them, M is the number of antennas, N is the number of devices to be detected. In this communication system, the device to be detected can send data to the base station. When the base station receives the signal data sent by the device to be detected, it detects the sent data of all active devices from the received data signal, and then determines the activity level of all devices to be detected. The following will explain the method for determining the activity status detection of the device to be detected.

[0074] Embodiment 1

[0075] Please refer to Figure 1 , Figure 1 It is a flowchart of an embodiment of a device activity status detection method based on the PSCA algorithm provided by the present application, including steps S101 to S103;

[0076] Step S101, receiving pilot signals sent by a plurality of devices to be detected;

[0077] In some embodiments, when each device to be detected is activated, it will send a pilot signal to the base station according to a predetermined protocol. At this time, the base station will receive pilot signals sent by several devices to be detected.

[0078] It should be noted that before receiving the pilot signal Y sent by several devices to be detected, a specific pilot sequence with a length of L needs to be allocated to each device to be detected. At the same time, the base station will know the pilot sequence s of all devices to be detected n, since L<<N, the pilot sequence s of different devices to be detected n For simplicity, it is assumed that the time and frequency are completely synchronized, that is, the active devices are guaranteed to send the pilot signal Y on the same frequency band at the same time.

[0079] In some embodiments, the relevant formula of the pilot signal Y is:

[0080]

[0081] Where, Y is the pilot signal sent by all devices to be detected; n is the nth device to be detected, is the set of devices to be detected; α n is the active state of the nth device to be detected, α n ∈{0,1}, where α n =1 means the device to be detected n is in active state, α n =0 means the device n to be detected is in an inactive state; g n is the path loss of the channel between the nth device to be detected and the base station; s n h is the pilot signal sent by the nth device to be detected; n represents the small-scale fading coefficient of the channel between the base station and the device to be detected n in each coherent block, where where h n,m represents the small-scale fading coefficient of the channel between the mth antenna of the base station and the device to be detected n; Z is the additive Gaussian white noise to which the base station is subjected, All elements of Z follow the complex Gaussian distribution CN(0,σ 2 ); S is the pilot signal matrix sent by all devices to be detected; Γ is the active state matrix of the devices to be detected, W is the path loss matrix, where H is the unknown small-scale channel coefficient matrix of the base station, where All its elements follow independent and identically distributed complex Gaussian distribution

[0082] It should be noted that Y is the pilot signal sent by all the devices to be detected and received by the base station, which is a matrix with L rows and N columns; n is the pilot signal sent by the nth device to be detected, which is an L-dimensional vector; S is the pilot signal matrix sent by all devices to be detected, which is a matrix with L rows and N columns, where the nth column is s n .

[0083] It should be noted that the path loss of the channel between each device to be detected and the base station is known, so the subsequent calculation can be directly based on this, and there is no need to calculate the path loss by other methods.

[0084] Step S102, inputting the pilot signal into a preset PSCA algorithm, calculating a first covariance matrix corresponding to the pilot signal in each iteration, determining a corresponding inverse matrix based on the first covariance matrix, and calculating in parallel the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected according to the inverse matrix, iteratively calculating the active state estimation value according to a preset algorithm step size and the optimal solution, and outputting a target active state estimation value when a preset iteration termination condition is met, wherein the active state estimation optimization problem is obtained by approximating the initial active state estimation optimization problem after applying it to each of the devices to be detected;

[0085] In some embodiments, before the pilot signal Y is input into the preset PSCA algorithm, the parameters required by the PSCA algorithm need to be initialized, including the activity status α of all devices to be detected, that is, α (0) =0, preset algorithm step size, number of iterations k = 0, etc. At the same time, the path loss vector g, pilot signal matrix S and noise power σ should be stored in advance in the PSCA algorithm 2 And other parameters.

[0086] In some embodiments, the pilot signal Y is input into a preset PSCA algorithm to calculate a first covariance matrix corresponding to the device active variable in each iteration: Specifically, after receiving the pilot signal Y, the pilot signal Y is passed to the preset PSCA algorithm for processing. At this time, the PSCA algorithm will call the pre-stored parameters (including the path loss moment vector g, the pilot signal matrix S and the noise power σ 2 etc.) to calculate the first covariance matrix corresponding to the device active variables Among them, the first covariance matrix The calculation formula is: In the formula, is the first covariance matrix; S is the pilot signal matrix sent by all devices to be detected; Γ is the active state matrix of the device to be detected (Γ (k) represents the active state matrix of all devices at the kth iteration), W is the path loss matrix, where σ 2 is the noise power; I L is an L×L identity matrix; L is the pilot signal length.

[0087] In some embodiments, based on the first covariance matrix Determine the corresponding inverse matrix Specifically: From the first covariance matrix The real part and the imaginary part of the complex number are extracted from the covariance matrix, and the first covariance matrix is ​​calculated based on the real part and the imaginary part. The corresponding inverse matrix Among them, based on the first covariance matrix Determine the corresponding inverse matrix The calculation formula is as follows:

[0088]

[0089]

[0090]

[0091] In the formula, and Represent the real and imaginary parts of a complex number respectively; is the inverse matrix; is the first covariance matrix; i is an imaginary unit.

[0092] It should be noted that the above-mentioned first covariance matrix Determine the corresponding inverse matrix This can be achieved using the Frobenius inversion method.

[0093] In this way, by determining the corresponding inverse matrix based on the first covariance matrix, an inverse matrix that can accurately reflect the relationship between the device activity state and the pilot signal can be obtained, which facilitates the subsequent accurate determination of the target activity state estimation value.

[0094] It should be noted that, when detecting the activity state of a device, the present application may express the problem of detecting the activity state of all devices through different activity state estimation optimization problems, wherein the activity state estimation optimization problem may be but is not limited to the maximum likelihood estimation problem (MLE) and the maximum a posteriori estimation problem (MAPE).

[0095] In some embodiments, when the problem of device activity state detection is expressed by the maximum likelihood estimation problem (MLE), the optimal solution of the activity state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix, including: determining the corresponding second covariance matrix based on the pilot signal Y And based on the second covariance matrix The first covariance matrix and the inverse matrix Determine the first estimation problem of device active detection The first estimation problem is obtained by applying the initial maximum likelihood estimation problem to each of the devices to be detected and then approximating the initial maximum likelihood estimation problem; Solve and obtain the corresponding optimal solution

[0096] In some embodiments, the corresponding second covariance matrix is ​​determined based on the pilot signal Y The calculation formula is as follows: In the formula, is the second covariance matrix, M is the number of base station antennas, Y is the pilot signal, and L is the length.

[0097] In some embodiments, based on the second covariance matrix The first covariance matrix and the inverse matrix Determine the first estimation problem of device active detection Specifically: Since the pilot signal Y follows an independent and identical distribution Therefore, based on the second covariance matrix The first covariance matrix and the inverse matrix Determine the initial maximum likelihood estimation problem Afterwards, the initial maximum likelihood estimation problem is After being applied to each of the devices to be detected, the first estimation problem is obtained by approximation Afterwards, for the first estimation problem Solve and obtain the corresponding optimal solution

[0098] In some embodiments, the initial maximum likelihood estimation problem The relevant formula is: The likelihood function based on the pilot signal Y is

[0099]

[0100] In the formula, |·| represents the determinant of the matrix, tr(·) represents the trace of the matrix, and M is the number of base station antennas. Then, for the likelihood function p ML-K Taking the negative logarithm of (Y; α) (discarding the constant term that is independent of α), we can get

[0101]

[0102] Therefore, the initial maximum likelihood estimation problem can be determined as:

[0103] It should be noted that when solving the initial maximum likelihood estimation problem When , it can be approximated into N first estimation problems for solution, where the expression of the first estimation problem is:

[0104] It should be noted that the initial maximum likelihood estimation problem is a non-convex problem, and the first estimation problem It is a convex problem.

[0105] In some embodiments, the first estimation problem Solve and obtain the corresponding optimal solution The calculation formula is as follows:

[0106]

[0107]

[0108]

[0109]

[0110] In the formula, is the first estimation problem of the nth device to be detected, α n is the active state of the nth device to be detected, is the estimated value of the active state of the kth iteration; n is the nth device to be detected, is a collection of the devices to be detected; is the first covariance matrix; is the inverse matrix; is the second covariance matrix; g n is the path loss of the channel between the nth device to be detected and the base station; s n A pilot signal sent by the nth device to be detected; is the optimal solution to the first estimation problem.

[0111] In this way, by solving the first estimation problem, the optimal solution corresponding to the initial maximum likelihood estimation problem can be quickly and accurately calculated, thereby accurately determining the active state of the device.

[0112] In some embodiments, when the problem of device activity state detection is expressed by the maximum a posteriori estimation problem (MAPE), the optimal solution of the activity state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix. Comprising: determining a corresponding second covariance matrix based on the pilot signal Y Determine the corresponding probability distribution function f based on the active state distribution of the plurality of devices to be detected t (α), and based on the second covariance matrix and the probability distribution function f t(α) Determine the initial maximum a posteriori estimation problem Afterwards, the initial maximum a posteriori estimation problem After being applied to each of the devices to be detected, the second estimation problem is obtained by approximation Afterwards, for the first estimation problem Solve and obtain the corresponding optimal solution

[0113] In some embodiments, the initial maximum a posteriori estimation problem The relevant formula is: Determine the corresponding probability distribution function f based on the active state distribution of the plurality of devices to be detected t (α), considering the distribution of device activity α under the general model, the conditional probability distribution function of A under a given pilot signal Y can be determined as: MAP-K (Y,α)=p ML-K (Y;α)p A-t (α), where p ML-K (Y; α) is the likelihood function of receiving the pilot signal Y,

[0114]

[0115] p A-t (α) is the probability distribution function of the device activity random variable, which can be the probability distribution function p of the general model below: A-G (α) or the probability distribution function p of the independent model A-I (α), Among them, t∈{G,I} represents the general model G or independent model I of device activity. Therefore, the initial maximum likelihood estimation problem can be determined as

[0116] It should be noted that when the active state estimation optimization problem is the second estimation problem (MAPE), the device state α can be modeled as a certain unknown (that is, the general model G) or as a Bernoulli random variable The implementation of in order to incorporate prior knowledge (hereinafter referred to as independent model I). (1) In the general model G, the probability distribution function p A-G The calculation formula of (α) is as follows:

[0117]

[0118] In the formula, Ψ represents a non-empty subset of N, c ω Is a reflection of α n , the coefficient of correlation between n∈ω, α n is the active state of the nth device to be detected, is the set of devices to be detected. (2) In the independent model I, the device states A of all devices to be detected are independent Bernoulli random variables, so the corresponding probability distribution function p A-I (α) is:

[0119]

[0120] In the formula, represents the active probability of the device n to be detected, α n is the active state of the nth device to be detected. In the multivariate Bernoulli model, c ω ,ω∈Ψ can be estimated from the historical information of device activity. If the probability distribution function of A in any form is given, c ω ,ω∈Ψ can be directly calculated. According to the probability distribution function p of the independent model A-I (α) can get the active probability of each device to be detected So when all |ω|>1, c ω =0, we can get: At this time, the probability distribution function p of the general model G A-G (α) can be simplified to the probability distribution function of the independent model, where

[0121] In some embodiments, based on the second covariance matrix and the probability distribution function f t (α) Determining the Second Estimation Problem for Device Activity Detection The expression of the initial maximum a posteriori estimation problem is: In solving the initial maximum a posteriori estimation problem When , it can be approximated into N second estimation problems for solution, where the expression of the second estimation problem is:

[0122] In some embodiments, the second estimation problem is solved to obtain the corresponding optimal solution. The calculation formula is as follows:

[0123]

[0124]

[0125]

[0126]

[0127]

[0128] In the formula, is the second estimation problem of the nth device to be detected; α n is the active state of the nth device to be detected, α (k) is the estimated value of the active state of the kth iteration; n is the nth device to be detected, is a collection of the devices to be detected; f is the kth iteration MAP-K (α (k) ) Regarding the active state α of the nth device to be detected n The gradient of α is the first covariance matrix; is the inverse matrix; is the second covariance matrix; f t ′(α (k) ) is the gradient of the negative logarithm of the probability distribution function of the device activity variable with respect to the activity state α; c ω Is active state α n The coefficient of correlation, n∈ω, ω∈Ψ, Ψ represents A non-empty subset of n represents the activity probability of the nth device to be detected, where G represents the index of the general model of device activity; I represents the index of the independent model of device activity; g n is the path loss of the channel between the nth device to be detected and the base station; s n A pilot signal sent by the nth device to be detected; is the optimal solution to the second estimation problem.

[0129] It should be noted that the subsequent optimal solution may be the optimal solution to the first estimation problem It can also be the optimal solution to the second estimation problem

[0130] In this way, by solving the second estimation problem, the optimal solution corresponding to the initial maximum a posteriori estimation problem can be quickly and accurately calculated, thereby accurately determining the active state of the device.

[0131] In some embodiments, according to a preset algorithm step size The optimal solution Estimated value of active status Perform iterative calculations, and when the preset iteration termination conditions are met, output the target active state estimate. Specifically, when the optimal solution is obtained When, only the step size of the algorithm is needed. Enter the formula for iterative calculation. When the number of iterations meets the preset number of iterations, the algorithm will terminate and the target active state estimation value can be output.

[0132] It should be noted that the number of iterations K can be customized according to needs, or the number of iterations K can be obtained indirectly by setting the stopping condition of the iterative algorithm. Generally speaking, under the same conditions, the larger the number of iterations K, the lower the detection error rate and the longer the calculation time.

[0133] In some embodiments, the step length of the algorithm is preset. The optimal solution The calculation formula for iteratively calculating the activity state estimation value is as follows:

[0134]

[0135] In the formula, and are the estimated values ​​of active states at the k+1th and kth iterations, respectively; is the preset algorithm step size; is the optimal solution of the active state estimation optimization problem; n is the nth device to be detected, is a set of devices to be detected.

[0136] For easier understanding of step S102, please refer to Table 1, which is the overall process of the PSCA algorithm;

[0137]

[0138] In some embodiments, the preset algorithm step size includes a first algorithm step size that satisfies a preset step size condition or a second algorithm step size determined after neural network optimization, specifically: the first algorithm step size satisfies In the formula, is the step size of the first algorithm; k is the number of iterations; the step size of the second algorithm is to train the neural network model using the acquired sample data set in a supervised manner until the loss function is minimized, and the model parameters of the neural network model are determined.

[0139] It should be noted that, when the number of iterations k of the first algorithm tends to infinity, the active state estimation value α (k) Converge to the stationary point of the maximum approximate estimation problem of device activity detection, that is, the PSCA algorithm will find the optimal solution.

[0140] It should be noted that a neural network model can be constructed to determine the second algorithm step size, wherein one layer of the neural network model implements one iteration of the PSCA algorithm, the number of network layers is the number of iterations, the speed of reducing the device active detection error rate decreases with the increase of the number of network layers, and the calculation time increases linearly with the number of network layers. Through the sample data set (the sample covariance matrix of the received pilot signal Equipment large-scale fading matrix W [i] and the known pilot sequence S [i] ) Train the neural network model to get the prediction results And calculate the actual active state of the device α [i] And the prediction results The loss function between , and continuously optimize the neural network model with the minimum loss value as the goal, so as to obtain the trained target neural network model PSCANet, where the network architecture diagram of the target neural network model is as follows Figure 2 As shown, the second algorithm step size can be determined at this time, where the loss value calculation formula is:

[0141] In the formula, is the prediction result, α [i] is the actual active status of the device. is the training sample set, α n is the active state of the nth device to be detected, and They represent the training sample set, the verification sample set and the test sample set respectively.

[0142] It should be noted that the loss function can also be binary cross entropy (BCE) and the like, which is not limited in this application.

[0143] S103: Obtaining an activity status detection result of each of the devices to be detected based on the target activity status estimation value.

[0144] In some embodiments, when the target activity state estimate is obtained After that, it needs to be compared with the preset threshold value θ. If it is greater than the preset threshold value θ, the device activity state can be set to 1, which means that the device to be detected is in an active state. Otherwise, the device activity state is set to 0, which means that the device to be detected is in an inactive state. The relevant formula is

[0145] It should be noted that the preset threshold can be obtained by minimizing the error between the actual active state of the sample data and the estimated active state, that is, It can be obtained or set freely, and this application does not impose any restrictions.

[0146] It should be noted that, for situations where the device activity distribution is more complex, a first-order or second-order approximate model of the device activity distribution can be used to simplify the device activity distribution, and then the method of the present application can be used to perform activity status detection.

[0147] It should be noted that the above algorithm flow can be deployed on hardware suitable for parallel computing (such as GPU, FPGA, DSP, etc.) for rapid implementation, and this application does not impose any restrictions.

[0148] The embodiment of the present application receives pilot signals sent by several devices to be detected, so as to facilitate the subsequent input of a preset PSCA algorithm to accurately and quickly detect the active state of the corresponding device; the pilot signal is input into the preset PSCA algorithm, and the corresponding inverse matrix is ​​determined based on the first covariance matrix, so as to obtain an inverse matrix that can accurately reflect the relationship between the active state of the device and the pilot signal, so as to facilitate the subsequent accurate determination of the target active state estimate; the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix, so as to quickly and accurately calculate the optimal solution corresponding to the active state estimation optimization problem, and then accurately determine the active state of the device; at the same time, the convergence speed and detection accuracy can be balanced by dynamically adjusting the algorithm step size. Therefore, compared with the prior art, the present application can significantly improve the accuracy and speed of device active state detection in the unauthorized access scenario.

[0149] Embodiment 2

[0150] Please refer to Figure 3 , Figure 3 It is a structural diagram of an embodiment of a device active state detection system based on a PSCA algorithm provided by the present application, comprising a receiving module 100, a processing module 200 and a determining module 300;

[0151] The receiving module 100 is used to receive pilot signals sent by a plurality of devices to be detected;

[0152] The processing module 200 is used to input the pilot signal into a preset PSCA algorithm, calculate a first covariance matrix corresponding to the pilot signal in each iteration, determine a corresponding inverse matrix based on the first covariance matrix, and calculate the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected in parallel according to the inverse matrix, iteratively calculate the active state estimation value according to a preset algorithm step size and the optimal solution, and output a target active state estimation value when a preset iteration termination condition is met, wherein the active state estimation optimization problem is obtained by approximating the initial active state estimation optimization problem after applying it to each of the devices to be detected;

[0153] The determination module 300 is used to obtain the activity state detection result of each of the devices to be detected based on the target activity state estimation value.

[0154] Since the information interaction, execution process, etc. between the modules in the above-mentioned device activity status detection system based on the PSCA algorithm are based on the same concept as the embodiment of the device activity status detection method based on the PSCA algorithm in the first aspect of the present invention, the technical effects achieved are basically the same. For specific contents, please refer to the description in the first embodiment of the method of the present invention, and will not be repeated here.

[0155] The device embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separated, i.e., may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the method of this embodiment.

[0156] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-monitorable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0157] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application.

[0158] It is particularly pointed out that for those skilled in the art, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A device activity status detection method based on PSCA algorithm, characterized in that: include: Receiving pilot signals sent by several devices to be detected; Input the pilot signal into a preset PSCA algorithm, calculate a first covariance matrix corresponding to the pilot signal in each iteration, determine a corresponding inverse matrix based on the first covariance matrix, and calculate the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected in parallel according to the inverse matrix, iteratively calculate the active state estimation value according to a preset algorithm step size and the optimal solution, and output a target active state estimation value when a preset iteration termination condition is met, wherein the active state estimation optimization problem is obtained by approximating the initial active state estimation optimization problem after applying it to each of the devices to be detected; Based on the target activity state estimation value, an activity state detection result of each of the devices to be detected is obtained.

2. The device activity status detection method based on the PSCA algorithm according to claim 1 is characterized in that: The determining of the corresponding inverse matrix based on the first covariance matrix is ​​specifically: The real part and the imaginary part of the complex number are extracted from the first covariance matrix, and an inverse matrix corresponding to the first covariance matrix is ​​calculated based on the real part and the imaginary part.

3. The device activity status detection method based on the PSCA algorithm according to claim 2 is characterized in that: The calculation formula for determining the corresponding inverse matrix based on the first covariance matrix is ​​specifically: In the formula, and Represent the real and imaginary parts of a complex number respectively; is the inverse matrix; is the first covariance matrix; i is an imaginary unit.

4. The device activity status detection method based on the PSCA algorithm according to claim 1 is characterized in that: The active state estimation optimization problem includes a first estimation problem, wherein the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix, specifically: Determine a corresponding second covariance matrix based on the pilot signal, and determine the first estimation problem for device activity state detection based on the second covariance matrix, the first covariance matrix and the inverse matrix, wherein the first estimation problem is approximated by applying an initial maximum likelihood estimation problem to each of the devices to be detected; The first estimation problem is solved to obtain the corresponding optimal solution.

5. The device activity status detection method based on the PSCA algorithm according to claim 4 is characterized in that: The first estimation problem is solved to obtain the corresponding calculation formula of the optimal solution, which is specifically: In the formula, is the first estimation problem of the nth device to be detected, α n is the active state of the nth device to be detected, is the estimated value of the active state of the kth iteration; n is the nth device to be detected, is a collection of the devices to be detected; is the first covariance matrix; is the inverse matrix; is the second covariance matrix; g n is the path loss of the channel between the nth device to be detected and the base station; s n A pilot signal sent by the nth device to be detected; is the optimal solution to the first estimation problem.

6. The device activity status detection method based on the PSCA algorithm according to claim 1 is characterized in that: The active state estimation optimization problem also includes a second estimation problem, wherein the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected is calculated in parallel according to the inverse matrix, specifically: Determine a corresponding second covariance matrix based on the pilot signal; Determine a corresponding probability distribution function based on the activity state distribution of the plurality of devices to be detected, and determine the second estimation problem for device activity state detection based on the second covariance matrix and the probability distribution function, wherein the second estimation problem is obtained by approximating an initial maximum a posteriori estimation problem after applying it to each device to be detected; The second estimation problem is solved to obtain the corresponding optimal solution.

7. The device activity status detection method based on the PSCA algorithm according to claim 6 is characterized in that: The second estimation problem is solved to obtain the corresponding calculation formula of the optimal solution, which is specifically: In the formula, is the second estimation problem of the nth device to be detected; α n is the active state of the nth device to be detected, α (k) is the estimated value of the active state of the kth iteration; n is the nth device to be detected, is a collection of the devices to be detected; is the first covariance matrix; is the inverse matrix; is the second covariance matrix; is the gradient of the negative logarithm of the probability distribution function of the active state at the kth iteration with respect to the active state α; c ω Is active state α n The coefficient of correlation, n∈ω, ω∈Ψ, Ψ represents A non-empty subset of n represents the activity probability of the nth device to be detected, where G represents the index of the general model of device activity distribution; I represents the index of the independent model of device activity distribution; g n is the path loss of the channel between the nth device to be detected and the base station; s n A pilot signal sent by the nth device to be detected; is the optimal solution to the second estimation problem.

8. The device activity status detection method based on the PSCA algorithm according to claim 1, characterized in that: The calculation formula for iteratively calculating the activity state estimation value according to the preset algorithm step size and the optimal solution is specifically: In the formula, and are the estimated values ​​of active states at the k+1th and kth iterations, respectively; is the preset algorithm step size; is the optimal solution of the active state estimation optimization problem; n is the nth device to be detected, is a set of devices to be detected.

9. The device activity status detection method based on the PSCA algorithm according to claim 1, characterized in that: The preset algorithm step size includes a first algorithm step size that meets the preset step size condition or a second algorithm step size determined after neural network optimization, specifically: The first algorithm step size satisfies In the formula, is the step size of the first algorithm; k is the number of iterations; The second algorithm step size is to train the neural network model using the acquired sample data set in a supervised manner until the loss function is minimized, and the model parameters of the neural network model are determined.

10. A device activity status detection system based on PSCA algorithm, characterized in that: include: A receiving module, a processing module and a determining module; The receiving module is used to receive pilot signals sent by a plurality of devices to be detected; The processing module is used to input the pilot signal into a preset PSCA algorithm, calculate a first covariance matrix corresponding to the pilot signal in each iteration, determine a corresponding inverse matrix based on the first covariance matrix, and calculate in parallel the optimal solution of the active state estimation optimization problem corresponding to each of the devices to be detected according to the inverse matrix, iteratively calculate the active state estimation value according to a preset algorithm step size and the optimal solution, and output a target active state estimation value when a preset iteration termination condition is met, wherein the active state estimation optimization problem is obtained by approximating the initial active state estimation optimization problem after applying it to each of the devices to be detected; The determination module is used to obtain the activity state detection result of each of the devices to be detected based on the target activity state estimation value.

Images