CN115865296B - Orthogonal pilot sequence active detection method based on covariance - Google Patents

Abstract

The invention discloses an orthogonal pilot sequence active detection method based on covariance. The invention solves the problem of active user detection in a large-scale unlicensed access in a cellular-free distributed large-scale MIMO communication system, and breaks through the bottleneck problem of the traditional active user detection algorithm affected by various interferences. The invention firstly provides a covariance interference measurement mode among devices in a communication system, and determines pilot sequences and transmitting power sent by all access devices in the communication system by using a method for allocating minimum-maximum covariance interference pilot. The present invention then employs a partial update coordinate descent method to detect device activity in pilot allocation mode of minimum-maximum covariance interference. The method avoids the problem of overlarge pilot overhead in a non-orthogonal pilot sequence detection algorithm, and has very important significance for processing the problem of active user detection in a mobile scene, so the method has certain practical value.

Description

Orthogonal pilot sequence active detection method based on covariance

Technical Field

The invention relates to the technical field of active user detection in a honeycomb-free large-scale MIMO system, in particular to a covariance-based orthogonal pilot sequence active detection scheme using a minimum-maximum covariance interference (Min-Max Covariance Interference, MMCI) pilot allocation method and a partial update coordinate descent method in large-scale unlicensed access of the honeycomb-free large-scale MIMO system.

Background

With the rapid development of large-scale internet of things (MASSIVE INTERNET of Things, mIoT), large-scale machine type communication and ultra-reliable low-latency communication requirements in the 5G standard are combined into new 6G services, and become large-scale ultra-reliable low-latency communication (massive Ultra Reliable Low Latency Communication, mURLLC) requirements. To guarantee low latency requirements for large-scale access to latency sensitive services, unlicensed random access (GFRA) techniques are proposed to reduce access latency and signaling overhead. The primary challenge faced by unlicensed random access technology is active user detection. mIoT, although there are a large number of potential devices connected to the network, these devices do not have service requirements at all times, and only a small percentage of the devices are active at a given time. For ease of identification, each device is pre-assigned a pilot sequence when accessing the network. However, due to the limited time-frequency resources, it is not possible to assign each device with mutually orthogonal pilot sequences, but the mutual interference between devices using non-orthogonal pilot sequences is not negligible. Therefore, an effective scheme needs to be designed, the limit of the limited orthogonal pilot sequence length on the activity detection of the devices is overcome, and the pilot interference among the devices is reduced, so that the accuracy of the active user detection of mURLLC devices in mIoT networks is improved.

Disclosure of Invention

Technical problems: in view of the above, the present invention aims to provide a covariance-based scheme for active detection (Orthogonal Pilot Sequences based Activity Detection, OPSAD) of orthogonal pilot sequences, which combines the characteristics of a non-cellular distributed massive MIMO system with good cross-correlation properties of orthogonal pilot sequences, so as to realize high-precision and low-complexity active device detection. The invention designs a method for pilot frequency distribution by using minimum-maximum covariance interference, which determines pilot frequency sequences transmitted by each access device, power of transmitting pilot frequency and AP cluster for detecting the active state of the device in a communication system, and then adopts a scheme for detecting the device activity in the pilot frequency distribution mode by adopting a partial update coordinate descent method.

The technical scheme is as follows: the invention relates to an orthogonal pilot sequence active detection method based on covariance, which specifically comprises the following steps:

step1, establishing a channel model of a honeycomb-free large-scale MIMO system, and obtaining an expression of a received signal covariance matrix for active user detection;

Step 2, defining new inter-equipment interference measurement to calculate the influence of interference on active user detection, designing a pilot frequency distribution method of minimum-maximum covariance interference, determining a pilot frequency sequence transmitted by equipment with minimum inter-access equipment interference in a communication system, transmitting pilot frequency power and detecting an AP cluster of the active state of each equipment;

And 3, adopting a partial update coordinate descent method to detect the active users of the pilot frequency allocation algorithm which minimizes the maximum interference among devices in the covariance-based active user detection algorithm.

Wherein, the step 1 specifically includes:

Step 101: in a honeycomb-free large-scale MIMO system, M Access Points (AP) with N antennas are distributed randomly, and all the APs are connected to a Central Processing Unit (CPU) through a return link; the total equipment number in the system is K, and each equipment is assumed to be provided with an antenna; let k be any integer in the set [1, k ], the active state of the kth device is represented by an active identifier a _k, a _k has 0 or 1 two values, wherein α _k =1 represents that the kth device is in an active state, and there is a service requirement; α _k =0 means that the kth device is in an inactive state, with no service requirements; the total number of the orthogonal pilot sequences which can be allocated in the system is tau _p, the length of each pilot sequence is L, and the pilot allocated by the kth user is recorded as Because there are multiple users sharing limited orthogonal pilot frequency, set total number tau _p of orthogonal pilot frequency sequence < total equipment number K;

considering a block fading channel model, the length of each coherent block is τ, m is set as any integer in the set [1, m ], and the channel matrix between the kth device and the mth AP is expressed as: Where β _mk represents the large-scale fading coefficient of the channel between the kth device and the mth AP, let the large-scale fading coefficient β _mk change slowly and be assumed to be known by the CPU; h _mk represents a small-scale fast fading vector of the channel between the kth device and the mth AP, which obeys a multivariate cyclosymmetric complex gaussian distribution with an average value of 0 and a correlation matrix of I _N, and I _N represents an identity matrix of N rows and N columns; assuming that the orthogonal pilot sequences allocated to the users are normalized, when l is an integer which is not equal to k in any one of the sets [1, k ], if the kth device and the first device use the same pilot sequence, the conjugate of the pilot of the kth device is multiplied by the pilot of the first device to be 1, otherwise, the conjugate of the pilot of the kth device is multiplied by the pilot of the first device to be 0;

in each coherent time block, all devices are divided into any one of tau _p orthogonal pilot sequences, and the pilot sequence allocated to the kth device is expressed as Of all the devices, only a small part of the devices are in an active state, and the small part of the active devices simultaneously transmit pilot sequences pre-allocated to the devices to all the APs, and then the signal received by the mth AP is recorded as Y _m:

Wherein the superscript symbol (·) ^H denotes a transpose operation on the matrix, Is a pilot matrix, is a complex matrix of L rows and K columns,A pilot, D _a＝diag(a₁,a₂,…,a_K), representing the k-th device assigned to, is a diagonal matrix of the active state matrix of the device, a _k is the active identifier of the k-th device,For a transmission power matrix of a device transmitting pilots, ρ _k is the pilot transmission power of the kth device, G _m＝[g_m1,g_m2,…,g_mK represents the channel matrix of all devices to the mth AP, G _mk is the channel matrix between the kth device and the mth AP, W _m represents an additive complex gaussian white noise matrix, each term of W _m is independent and compliantSign symbolRepresenting a multivariate cyclosymmetric complex gaussian distribution with zero mean and correlation coefficient σ ², σ ² being the variance value of the channel noise;

Step 102: according to formula (1), signal Y _m＝[y_m1,…,y_mn,…,y_mN received by the mth AP, wherein Representing the signal received by the nth antenna of the mth AP, g _mnk represents the channel between the kth device and the nth antenna of the mth AP, the channels received on the different antennas being independent, so the covariance matrix Q _m of Y _m can be defined as:

Wherein I _L represents an identity matrix of L rows and L columns, and a multivariate circular symmetric complex Gaussian distribution with y _mn obeying an average value of 0 and a correlation matrix of Q _m is obtained according to the definition of covariance.

The step 2 specifically includes:

Step 201: suppressing inter-device interference using the same pilot by AP selection and power control;

based on each received signal, covariance samples can be obtained Having sufficient statistical characteristics for active devices; the activity detection for the kth device in the mth AP is based onSignal component in (a)Because of the repeated use of orthogonal pilots, the signal good components transmitted by devices with the same pilots are indistinguishable in the covariance samples; using different sets of APs to detect active information of different devices by utilizing macro diversity gain and sparsity of power domains of a non-cellular massive MIMO system;

recording the AP set for detecting the activity of the kth device as Along withThe macro diversity gain of the kth device will increase, including an increase in the number of APs, but the interference using the same pilot will also increase, thus selecting for each deviceCan inhibit certain interference among devicesThe pilot signals received by all APs are expressed asThe covariance is expressed asIf the interference signal of other devices using the same pilot is negligibly weak in each device's serving AP cluster, then based onWill become more accurate;

Because covariance information collected by APs far away from the kth device has limited accuracy improvement for active detection of the kth device, and covariance of the pilot signal received by the AP with optimal channel conditions for the kth device can ensure accurate active detection of the kth device; so a master AP is selected for each device to detect the activity of that device, and the index of the master AP for the kth device is selected by equation (3):

Wherein the method comprises the steps of A value indicating that m is found that maximizes β _mk;

in order to further reduce inter-device interference, pilot transmission power control is introduced for the devices, and the transmission power of the kth device is selected as follows:

Where ρ _ds is the reception channel power expected by the master AP of the kth device, and the physical meaning of equation (4) is set to be the same for all devices: if the k-th device gets its master AP-index number as The channel condition of the AP of (2) is good enough, the kth device does not need to take too much transmit power; otherwise, the lack of channel conditions is compensated by using a larger transmitting power ρ _k; if ρ _k calculated by equation (4) exceeds the maximum transmit power that the kth device can provide, the kth device cannot access the network because its transmitted signal cannot be received by any AP identification;

step 202: defining a new inter-device interference metric to effectively calculate the impact of interference on active user detection;

Based on the AP selection and power control in the previous step, the inter-device interference in the covariance detection algorithm is further reduced by an allocation method, and as can be seen from equation (2), if the kth device and the first device allocate the same pilot, their signal components are indistinguishable at their respective main APs, wherein the interference level of the first device to the kth device is available Where ρ _l and ρ _k represent pilot transmit powers of the first and kth devices, respectively,Representing the first device and the first deviceLarge scale fading coefficients of channels between APs,Representing the kth device and the kth deviceLarge-scale fading coefficients of channels between APs;

Thus, defining the interference ζ _kl between the first device and the kth device is defined as:

Wherein equation (b) is derived from ζ _kl＝ζ_lk, i.e., the interference of the kth device to the first device is equal to the interference of the first device to the kth device;

Step 203: a pilot allocation algorithm that minimizes the maximum interference between devices in the covariance-based active user detection algorithm,

The number of APs in a large-scale high-reliability low-delay scene is far greater than the total number K of devices, at this time, the interference of pilot multiplexing is inevitably generated, but if each device and the adjacent devices allocate medium orthogonal pilots and the geographic positions of all devices using the same pilots are far apart, the maximum interference among all devices is guaranteed to be minimized, and the pilot allocation algorithm of the kth device is formulated as follows:

where f represents the maximum one of the devices with the same pilot as the kth device for the kth device, and the objective of this optimization problem is to let the kth device select the pilot F can be minimized;

To solve the problem of An algorithm for minimizing the maximum interference can be used, firstly, an interference diagram with weighted edges between all devices is established, and if two devices use the same pilot frequency, the weighted value between the devices is ζ _kl in the formula (5); if two devices use different pilots, the weighting value is 0, and then the pilot frequency distribution mode for minimizing the interference between the largest devices is found by iterating the condition of adopting each pilot frequency on each device, and the iteration is carried out until the pilot frequency matrix phi is not changed any more.

The step 3 specifically includes:

step 301: the active detection of the kth device can be formulated as a maximum likelihood problem;

The active user detection problem is essentially a maximum likelihood problem, when the active state of the kth device, a _k, is determined by Pilot signals received by all APsCalculating to obtain an activity state detection value of the kth deviceIf it isThe active user detects correctly, if not, the active user detects the error; the specific calculation process is as follows:

Since all large-scale fades are assumed to be known, then The likelihood function of (2) can be expressed as:

In the method, in the process of the invention, When the value denoted as a _k is determinedIs a function of the probability distribution of (1),Representation pair matrixThe sum of its diagonal elements is taken,Representing the base as natural constant e and the index asI pi Q _m represents a value of determinant which is obtained by multiplying the circumferential rate pi by Q _m to form a new matrix;

Wherein the active indications of devices other than the kth device are ignored and considered constant, then the maximum value of equation (10) is equivalent to The active user detection problem for the kth device can therefore be expressed as:

I.e. under the constraint of a _k E {0,1}, find the cause The minimum valueValues of Φ, a _k;

step 302: a partial update coordinate descent method is employed to simplify the scheme for detecting device activity,

According to the minimum-maximum covariance interference pilot allocation algorithm, each device is only actively detected by its main AP, and the pilot matrix phi adopted by each device is also selected; for the kth device, its generalized active state identifier Fu _k＝a_kρ_k is defined, and the generalized active state identifier detection value obtained when the active user detection algorithm calculates is recorded asThenThe active user detection problem of (2) can be reduced to:

I.e. at Under the constraint of this condition, find the causeThe minimum valueIs a value of (2); the coordinate descent method can be used for solvingIs to (1) optimize the problem ofFirst initialize to matrix σ ² I, where I represents the identity matrix, and use the current one in each iterationSolving forObtainingAnd then solving to obtainUpdatingThe iterative process is explained in detail below;

setting the maximum iteration number as T, taking T as any integer in the set [1, T ], and detecting the generalized active target state identifier detection value of the kth device calculated in the T-1 th iteration process Is marked asDetecting the calculated generalized active mark state identifier in the t-th iteration processIs marked asAnd iterate out t-1Obtained by adding the optimal updated change value d _t calculated in the t-th iteration processI.e.Therefore, the best updated change value d _t of each iteration needs to be found so that the iteration can be performedApproaching the actual value more and more; the solution process of d _t is as follows:

firstly, using Sherman-Morrison updating algorithm, Can be updated as:

Wherein the method comprises the steps of Is a computable number such thatThe optimization problem of (c) can be rewritten as:

I.e. at Under the constraint of this condition, find the minimum value of c _k+loge_k-b_k Is a value of (2); wherein the method comprises the steps ofIs a constant independent of d _t,Is a number associated with d _t;

Thus deriving loge _k-b_k and finding a value that lets it equal 0, the best update change value d _t in the t-th iteration can be solved for

To ensure thatAccording to the non-negativity of (2)Calculated outIs a negative number, and a heavy guarantee is added to d _t: I.e. if D _t takesIs then passed throughObtaining the calculated in the t-th iteration processThus, the t-th iteration ends; then starting the t+1st iteration until the maximum iteration number T is reached,

According to the best obtained after the end of T iterationsActive state detection value for kth deviceCan be obtained by (15):

Wherein the method comprises the steps of Is the threshold for active user detection of the kth device whenWhen the active state detection value for the kth deviceTake a value of 1, whenIn the time-course of which the first and second contact surfaces,The value is 0; if the active state detection value of the kth deviceEqual to the active identifier a _k of the kth device, the active user detects correct and if not, detects errors.

The beneficial effects are that: the invention provides an orthogonal pilot sequence active detection method based on covariance. Aiming at the problem of active user detection in the large-scale unlicensed access in the non-cellular distributed large-scale MIMO communication system, the power domain sparsity of the non-cellular distributed large-scale MIMO system is fully utilized while pilot frequency overhead is reduced, and the bottleneck problem of the traditional active user detection algorithm affected by multiple interferences is broken through. The invention firstly provides a covariance interference measurement mode between devices in a communication system, and determines a pilot sequence sent by each access device, power of sending pilot and an AP cluster for detecting the active state of the devices in the communication system by using a method for allocating minimum-maximum covariance interference (MMCI) pilot. The present invention then employs a partial update coordinate descent method to detect device activity in pilot allocation mode of minimum-maximum inter-device covariance interference. Compared with the traditional covariance-based active user detection method, the method provided by the invention reduces the interference among devices by AP clustering, power control and pilot frequency distribution, improves the accuracy of a detection algorithm, has low calculation complexity and high calculation speed by using a partial update coordinate reduction method, and has very important significance for processing the active user detection problem in a mobile scene.

Drawings

Fig. 1 is an overall flow chart of a covariance-based Orthogonal Pilot Sequence Activity Detection (OPSAD) algorithm;

Fig. 2 shows a comparison of the performance of 4 different active user detection methods under a given parameter configuration, each point representing the result of one detection threshold, and the corresponding false alarm probabilities and false miss probabilities are shown in the abscissa and ordinate, respectively. Wherein, CAD represents an active user detection algorithm based on clustering, and I represents the number of APs in each AP cluster; E-OPSAD represents an enhanced OPSAD algorithm in which the active state identifier of each device is used to update the covariance matrix of all access points; S-CAD represents a simplified CAD algorithm in which the active state identifier of each device is used to update the covariance matrix of its assigned access point;

FIG. 3 shows the trend of performance of OPSAD algorithm and CAD algorithm with the number of antennas per AP, the number of APs and the signal-to-noise ratio SNR value;

fig. 4 shows the effect of different pilot sequence lengths L and signal-to-noise ratio SNR on OPSAD algorithm and CAD algorithm.

Detailed Description

The invention is described in detail below with reference to examples:

Assuming a non-cellular distributed massive MIMO scenario, there are m=800 APs in the scenario, each AP is equipped with N antennas, and there are k=400 devices in the scenario. All APs and devices are randomly distributed in an area with radius r=0.5 km, and the large-scale fading coefficient is set as Where d _mk denotes the distance between the mth AP and the kth device, F _mk denotes the shadow fading component, which obeys a mean value of 0 and a correlation coefficient ofA multi-variable cyclic symmetric complex gaussian distribution,Representing the shadow fading variance value.

The number of symbols in one coherent time block of the system is τ=200, and the noise variance is σ ² = -109dBm. The orthogonal pilot sequences are generated by a Discrete Fourier Transform (DFT) matrix in the Galois field, the non-orthogonal pilot sequences are generated by a random complex gaussian matrix whose elements satisfy a multivariate circularly symmetric complex gaussian distribution with a distribution mean of 0 and a correlation coefficient of 1.

Based on the above-provided non-cellular distributed massive MIMO scenario, the covariance-based Orthogonal Pilot Sequence Activity Detection (OPSAD) scheme provided in this example specifically includes the following steps:

and step 1, establishing a channel model of the honeycomb-free large-scale MIMO system to obtain an expression of a received signal covariance matrix for active user detection.

In this example, step 1 specifically includes:

Step 101: considering a block fading channel model, each coherent block has a length τ, and the channel matrix between the kth device and the mth AP is expressed as: Where β _mk represents the large-scale fading coefficient of the channel between the kth device and the mth AP, which can be determined by Calculating; h _mk represents a small-scale fast fading vector of the channel between the kth device and the mth AP, which obeys a multivariate cyclosymmetric complex gaussian distribution with an average value of 0 and a correlation matrix of I _N, and I _N represents an identity matrix of N rows and N columns; assuming that the orthogonal pilot sequences allocated to the users are normalized, when l is an integer which is not equal to k in any one of the sets [1, k ], if the kth device and the first device use the same pilot sequence, the conjugate of the pilot of the kth device is multiplied by the pilot of the first device to be 1, otherwise, the conjugate of the pilot of the kth device is multiplied by the pilot of the first device to be 0;

in each coherent time block, all devices are divided into any one of tau _p orthogonal pilot sequences, and the pilot sequence allocated to the kth device is expressed as Of all the devices, only a small part of the devices are in an active state, and the small part of the active devices simultaneously transmit pilot sequences pre-allocated to the devices to all the APs, and then the signal received by the mth AP is recorded as Y _m:

Wherein the superscript symbol (·) ^H denotes a transpose operation on the matrix, Is a pilot matrix, is a complex matrix of L rows and K columns,A pilot, D _a＝diag(a₁,a₂,…,a_K), representing the k-th device assigned to, is a diagonal matrix of the active state matrix of the device, a _k is the active identifier of the k-th device,For a transmission power matrix of a device transmitting pilots, ρ _k is the pilot transmission power of the kth device, G _m＝[g_m1,g_m2,…,g_mK represents the channel matrix of all devices to the mth AP, G _mk is the channel matrix between the kth device and the mth AP, W _m represents an additive complex gaussian white noise matrix, each term of W _m is independent and compliantSign symbolRepresenting a multivariate cyclosymmetric complex gaussian distribution with zero mean and correlation coefficient σ ², σ ² being the variance value of the channel noise;

Step 102: according to formula (1), signal Y _m＝[y_m1,…,y_mn,…,y_mN received by the mth AP, wherein Representing the signal received by the nth antenna of the mth AP, g _mnk represents the channel between the kth device and the nth antenna of the mth AP, the channels received on the different antennas being independent, so the covariance matrix Q _m of Y _m can be defined as:

Wherein I _L represents an identity matrix of L rows and L columns, and a multivariate circular symmetric complex Gaussian distribution with y _mn obeying an average value of 0 and a correlation matrix of Q _m is obtained according to the definition of covariance.

And 2, defining a new inter-device interference metric to calculate the influence of interference on active user detection, designing a pilot frequency distribution method of minimum-maximum covariance interference, determining a pilot frequency sequence transmitted by a device with minimum inter-access device interference in a communication system, transmitting pilot frequency power and detecting an AP cluster of the active state of each device.

In this example, step 2 specifically includes:

Step 201: suppressing inter-device interference using the same pilot by AP selection and power control;

based on each received signal, covariance samples can be obtained Having sufficient statistical characteristics for active devices; the activity detection for the kth device in the mth AP is based onSignal component in (a)Because of the repeated use of orthogonal pilots, the signal good components transmitted by devices with the same pilots are indistinguishable in the covariance samples; using different sets of APs to detect active information of different devices by utilizing macro diversity gain and sparsity of power domains of a non-cellular massive MIMO system;

recording the AP set for detecting the activity of the kth device as Along withThe macro diversity gain of the kth device will increase, including an increase in the number of APs, but the interference using the same pilot will also increase, thus selecting for each deviceCan inhibit certain interference among devicesThe pilot signals received by all APs are expressed asThe covariance is expressed asIf the interference signal of other devices using the same pilot is negligibly weak in each device's serving AP cluster, then based onWill become more accurate;

Because covariance information collected by APs far away from the kth device has limited accuracy improvement for active detection of the kth device, and covariance of the pilot signal received by the AP with optimal channel conditions for the kth device can ensure accurate active detection of the kth device; so a master AP is selected for each device to detect the activity of that device, and the index of the master AP for the kth device is selected by equation (3):

Wherein the method comprises the steps of A value indicating that m is found that maximizes β _mk;

in order to further reduce inter-device interference, pilot transmission power control is introduced for the devices, and the transmission power of the kth device is selected as follows:

Where ρ _ds is the reception channel power expected by the master AP of the kth device, and the physical meaning of equation (4) is set to be the same for all devices: if the k-th device gets its master AP-index number as The channel condition of the AP of (2) is good enough, the kth device does not need to take too much transmit power; otherwise, the lack of channel conditions is compensated by using a larger transmitting power ρ _k; if ρ _k calculated by equation (4) exceeds the maximum transmit power that the kth device can provide, the kth device cannot access the network because its transmitted signal cannot be received by any AP identification;

step 202: defining a new inter-device interference metric to effectively calculate the impact of interference on active user detection;

Based on the AP selection and power control in the previous step, the inter-device interference in the covariance detection algorithm is further reduced by an allocation method, and as can be seen from equation (2), if the kth device and the first device allocate the same pilot, their signal components are indistinguishable at their respective main APs, wherein the interference level of the first device to the kth device is available Where ρ _l and ρ _k represent pilot transmit powers of the first and kth devices, respectively,Representing the first device and the first deviceLarge scale fading coefficients of channels between APs,Representing the kth device and the kth deviceLarge-scale fading coefficients of channels between APs;

Thus, defining the interference ζ _kl between the first device and the kth device is defined as:

Wherein equation (b) is derived from ζ _kl＝ζ_lk, i.e., the interference of the kth device to the first device is equal to the interference of the first device to the kth device;

Step 203: a pilot allocation algorithm that minimizes the maximum interference between devices in the covariance-based active user detection algorithm,

The number of APs in a large-scale high-reliability low-delay scene is far greater than the total number K of devices, at this time, the interference of pilot multiplexing is inevitably generated, but if each device and the adjacent devices allocate medium orthogonal pilots and the geographic positions of all devices using the same pilots are far apart, the maximum interference among all devices is guaranteed to be minimized, and the pilot allocation algorithm of the kth device is formulated as follows:

where f represents the maximum one of the devices with the same pilot as the kth device for the kth device, and the objective of this optimization problem is to let the kth device select the pilot F can be minimized;

To solve the problem of An algorithm for minimizing the maximum interference can be used, firstly, an interference diagram with weighted edges between all devices is established, and if two devices use the same pilot frequency, the weighted value between the devices is ζ _kl in the formula (5); if two devices use different pilots, the weighting value is 0, and then the pilot frequency distribution mode for minimizing the interference between the largest devices is found by iterating the condition of adopting each pilot frequency on each device, and the iteration is carried out until the pilot frequency matrix phi is not changed any more.

And 3, adopting a partial update coordinate descent method to detect the active users of the pilot frequency allocation algorithm which minimizes the maximum interference among devices in the covariance-based active user detection algorithm.

In this example, step3 specifically includes:

step 301: the active detection of the kth device can be formulated as a maximum likelihood problem;

The active user detection problem is essentially a maximum likelihood problem, when the active state of the kth device, a _k, is determined by Pilot signals received by all APsCalculating to obtain an activity state detection value of the kth deviceIf it isThe active user detects correctly, if not, the active user detects the error; the specific calculation process is as follows:

Since all large-scale fades are assumed to be known, then The likelihood function of (2) can be expressed as:

In the method, in the process of the invention, When the value denoted as a _k is determinedIs a function of the probability distribution of (1),Representation pair matrixThe sum of its diagonal elements is taken,Representing the base as natural constant e and the index asI pi Q _m represents a value of determinant which is obtained by multiplying the circumferential rate pi by Q _m to form a new matrix;

Wherein the active indications of devices other than the kth device are ignored and considered constant, then the maximum value of equation (10) is equivalent to The active user detection problem for the kth device can therefore be expressed as:

I.e. under the constraint of a _k E {0,1}, find the cause The minimum valueA value of D _ρ、Φ、a_k;

step 302: a partial update coordinate descent method is employed to simplify the scheme for detecting device activity,

According to the minimum-maximum covariance interference pilot allocation algorithm, each device is only actively detected by its main AP, and the pilot matrix phi adopted by each device is also selected; for the kth device, its generalized active state identifier Fu _k＝a_kρ_k is defined, and the generalized active state identifier detection value obtained when the active user detection algorithm calculates is recorded asThenThe active user detection problem of (2) can be reduced to:

I.e. at Under the constraint of this condition, find the causeThe minimum valueIs a value of (2); the coordinate descent method can be used for solvingIs to (1) optimize the problem ofFirst initialize to matrix σ ² I, where I represents the identity matrix, and use the current one in each iterationSolving forObtainingAnd then solving to obtainUpdatingThe iterative process is explained in detail below;

setting the maximum iteration number as T, taking T as any integer in the set [1, T ], and detecting the generalized active target state identifier detection value of the kth device calculated in the T-1 th iteration process Is marked asDetecting the calculated generalized active mark state identifier in the t-th iteration processIs marked asAnd iterate out t-1Obtained by adding the optimal updated change value d _t calculated in the t-th iteration processI.e.Therefore, the best updated change value d _t of each iteration needs to be found so that the iteration can be performedApproaching the actual value more and more; the solution process of d _t is as follows:

firstly, using Sherman-Morrison updating algorithm, Can be updated as:

Wherein the method comprises the steps of Is a computable number such thatThe optimization problem of (c) can be rewritten as:

I.e. at Under the constraint of this condition, find the minimum value of c _k+loge_k-b_k Is a value of (2); wherein the method comprises the steps ofIs a constant independent of d _t,Is a number associated with d _t;

Thus deriving loge _k-b_k and finding a value that lets it equal 0, the best update change value d _t in the t-th iteration can be solved for

To ensure thatAccording to the non-negativity of (2)Calculated outIs a negative number, and a heavy guarantee is added to d _t: I.e. if D _t takesIs then passed throughObtaining the calculated in the t-th iteration processThus, the t-th iteration ends; then starting the t+1st iteration until the maximum iteration number T is reached,

According to the best obtained after the end of T iterationsActive state detection value for kth deviceCan be obtained by (15):

Wherein the method comprises the steps of Is the threshold for active user detection of the kth device whenWhen the active state detection value for the kth deviceTake a value of 1, whenIn the time-course of which the first and second contact surfaces,The value is 0; if the active state detection value of the kth deviceEqual to the active identifier a _k of the kth device, the active user detects correct and if not, detects errors.

The above presents the whole process based on active user detection in a cellular-free distributed massive MIMO system using the method provided by this example.

FIG. 1 is an overall flow chart of the OPSAD algorithm;

Figure 2 shows a comparison of the performance of 4 different active user detection methods for a given parameter configuration. Each point represents the result of a detection threshold, and the corresponding false alarm probability and miss probability are displayed in the abscissa and ordinate respectively. Wherein, CAD represents an active user detection algorithm based on clustering, and I represents the number of APs in each AP cluster; E-OPSAD represents an enhanced OPSAD algorithm in which the active state identifier of each device is used to update the covariance matrix of all access points; the S-CAD represents a simplified CAD algorithm in which the active state identifier of each device is used to update the covariance matrix of its designated access point. It can be seen that the OPSAD algorithm outperforms the CAD algorithm when i=1, but the performance improvement of the CAD algorithm drops dramatically as I increases. This shows that a decrease in the update frequency of the covariance matrix estimated at the AP results in a significant decrease in the detection performance of CAD algorithms with non-orthogonal pilots, but has little impact on the detection performance with orthogonal pilots, demonstrating the effectiveness of the proposed OPSAD algorithm in reducing inter-device interference in covariance information at the AP;

Fig. 3 shows the trend of the performance of OPSAD algorithm and CAD algorithm with the number of antennas per AP, the number of APs and the SNR value. As the number of antennas increases, the performance of both algorithms improves significantly. However, when the signal-to-noise ratio is low (when the signal-to-noise ratio is less than or equal to 10 in the figure), the performance of the CAD algorithm is always poorer than that of the OPSAD algorithm under the same antenna configuration even if the number of access points is large;

Fig. 4 shows the effect of different pilot sequence lengths L and signal-to-noise ratio SNR on OPSAD algorithm and CAD algorithm. The error probability of both algorithms decreases rapidly with increasing signal-to-noise ratio, but the performance of OPSAD algorithm is always better than CAD algorithm, and OPSAD algorithm can reach the upper limit of performance with shorter pilot sequence than CAD algorithm to achieve the same detection performance.

In summary, the invention provides a covariance-based orthogonal pilot sequence activity detection method for the problem of active user detection in mURLLC scenes. The power domain sparsity of the honeycomb-free distributed large-scale MIMO system is fully utilized while pilot frequency overhead is reduced, and the bottleneck problem that the traditional active user detection algorithm is affected by multiple interferences is broken through. The method reduces the complexity of calculation while improving the detection precision, and has very important significance for processing the detection problem of the active user in the mobile scene, so the method has a certain practical value.

The present invention is not described in detail in the present application, and is well known to those skilled in the art.

The foregoing describes in detail preferred embodiments of the present invention. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the invention by one of ordinary skill in the art without undue burden. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by the person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.

Claims (1)

1. The covariance-based orthogonal pilot sequence activity detection method is characterized by comprising the following steps:

step1, establishing a channel model of a honeycomb-free large-scale MIMO system, and obtaining an expression of a received signal covariance matrix for active user detection;

Step 2, defining new inter-equipment interference measurement to calculate the influence of interference on active user detection, designing a pilot frequency distribution method of minimum-maximum covariance interference, determining a pilot frequency sequence transmitted by equipment with minimum inter-access equipment interference in a communication system, transmitting pilot frequency power and detecting an AP cluster of the active state of each equipment;

Step 3, adopting a partial update coordinate descent method to detect active users of a pilot frequency allocation algorithm which minimizes the maximum interference among devices in a covariance-based active user detection algorithm;

the step 1 specifically includes:

step 101: in a honeycomb-free large-scale MIMO system, M Access Points (AP) with N antennas are distributed randomly, and all the APs are connected to a Central Processing Unit (CPU) through a return link; the total equipment number in the system is K, and each equipment is assumed to be provided with an antenna; let k be any integer in the set [1, k ], the active state of the kth device is represented by an active identifier a _k, a _k has 0 or 1 two values, wherein a _k =1 represents that the kth device is in the active state, and there is a service requirement; a _k =0 indicates that the kth device is in an inactive state, with no service requirements; the total number of the orthogonal pilot sequences which can be allocated in the system is tau _p, the length of each pilot sequence is L, and the pilot allocated by the kth user is recorded as Because there are multiple users sharing limited orthogonal pilot frequency, set total number tau _p of orthogonal pilot frequency sequence < total equipment number K;

considering a block fading channel model, the length of each coherent block is τ, m is set as any integer in the set [1, m ], and the channel matrix between the kth device and the mth AP is expressed as: Where β _mk represents the large-scale fading coefficient of the channel between the kth device and the mth AP, let the large-scale fading coefficient β _mk change slowly and be assumed to be known by the CPU; h _mk represents a small-scale fast fading vector of the channel between the kth device and the mth AP, which obeys a multivariate cyclosymmetric complex gaussian distribution with an average value of 0 and a correlation matrix of I _N, and I _N represents an identity matrix of N rows and N columns; assuming that the orthogonal pilot sequences allocated to the users are normalized, when l is an integer which is not equal to k in any one of the sets [1, k ], if the kth device and the first device use the same pilot sequence, the conjugate of the pilot of the kth device is multiplied by the pilot of the first device to be 1, otherwise, the conjugate of the pilot of the kth device is multiplied by the pilot of the first device to be 0;

in each coherent time block, all devices are divided into any one of tau _p orthogonal pilot sequences, and the pilot sequence allocated to the kth device is expressed as Of all the devices, only a small part of the devices are in an active state, and the small part of the active devices simultaneously transmit pilot sequences pre-allocated to the devices to all the APs, and then the signal received by the mth AP is recorded as Y _m:

Wherein the superscript symbol (·) ^H denotes a transpose operation on the matrix, Is a pilot matrix, is a complex matrix of L rows and K columns,A pilot, D _a＝diag(a₁,a₂,…,a_K), representing the k-th device assigned to, is a diagonal matrix of the active state matrix of the device, a _k is the active identifier of the k-th device,For a transmission power matrix of a device transmitting pilots, ρ _k is the pilot transmission power of the kth device, G _m＝[g_m1,g_m2,…,g_mK represents the channel matrix of all devices to the mth AP, G _mk is the channel matrix between the kth device and the mth AP, W _m represents an additive complex gaussian white noise matrix, each term of W _m is independent and compliantSign symbolRepresenting a multivariate cyclosymmetric complex gaussian distribution with zero mean and correlation coefficient σ ², σ ² being the variance value of the channel noise;

Step 102: according to formula (1), signal Y _m＝[y_m1,…,y_mn,…,y_mN received by the mth AP, wherein Representing the signal received by the nth antenna of the mth AP, g _mnk represents the channel between the kth device and the nth antenna of the mth AP, the channels received on the different antennas being independent, so the covariance matrix Q _m of Y _m is defined as:

Wherein I _L represents an identity matrix of L rows and L columns, and a multivariate circular symmetric complex Gaussian distribution with y _mn obeying an average value of 0 and a correlation matrix of Q _m is obtained according to the definition of covariance;

the step 2 specifically includes:

Step 201: suppressing inter-device interference using the same pilot by AP selection and power control;

based on each received signal, covariance samples are obtained Having sufficient statistical characteristics for active devices; the activity detection for the kth device in the mth AP is based onSignal component in (a)Because of the repeated use of orthogonal pilots, the signal good components transmitted by devices with the same pilots are indistinguishable in the covariance samples; using different sets of APs to detect active information of different devices by utilizing macro diversity gain and sparsity of power domains of a non-cellular massive MIMO system;

recording the AP set for detecting the activity of the kth device as Along withThe macro diversity gain of the kth device will increase, including an increase in the number of APs, but the interference using the same pilot will also increase, thus selecting for each deviceCan inhibit certain interference among devicesThe pilot signals received by all APs are expressed asThe covariance is expressed asIf the interference signals of other devices using the same pilot are weak to negligible in each device's serving AP cluster, then based onWill become more accurate;

Because covariance information collected by APs far away from the kth device has limited accuracy improvement for active detection of the kth device, and covariance of the pilot signal received by the AP with optimal channel conditions for the kth device can ensure accurate active detection of the kth device; so a master AP is selected for each device to detect the activity of that device, and the index of the master AP for the kth device is selected by equation (3):

Wherein the method comprises the steps of A value indicating that m is found that maximizes β _mk;

in order to further reduce inter-device interference, pilot transmission power control is introduced for the devices, and the transmission power of the kth device is selected as follows:

Where ρ _ds is the reception channel power expected by the master AP of the kth device, and the physical meaning of equation (4) is set to be the same for all devices: if the k-th device gets its master AP-index number as The channel condition of the AP of (2) is good enough, the kth device does not need to take too much transmit power; otherwise, the lack of channel conditions is compensated by using a larger transmitting power ρ _k; if ρ _k calculated by equation (4) exceeds the maximum transmit power provided by the kth device, the kth device cannot access the network because its transmitted signal cannot be received by any AP identity;

step 202: defining a new inter-device interference metric to effectively calculate the impact of interference on active user detection;

Based on the AP selection and power control in the previous step, the inter-device interference in the covariance detection algorithm is further reduced by an allocation method, and as can be seen from the formula (2), if the kth device and the first device allocate the same pilot, their signal components are indistinguishable at their respective main APs, wherein the interference level of the first device to the kth device is available Where ρ _l and ρ _k represent pilot transmit powers of the first and kth devices, respectively,Representing the first device and the first deviceLarge scale fading coefficients of channels between APs,Representing the kth device and the kth deviceLarge-scale fading coefficients of channels between APs;

Thus, defining the interference ζ _kl between the first device and the kth device is defined as:

Wherein equation (b) is derived from ζ _kl＝ζ_lk, i.e., the interference of the kth device to the first device is equal to the interference of the first device to the kth device;

Step 203: a pilot allocation algorithm that minimizes the maximum interference between devices in the covariance-based active user detection algorithm,

The number of APs in a large-scale high-reliability low-delay scene is far greater than the total number K of devices, at this time, the interference of pilot multiplexing is inevitably generated, but if each device and the adjacent devices allocate medium orthogonal pilots and the geographic positions of all devices using the same pilots are far apart, the maximum interference among all devices is guaranteed to be minimized, and the pilot allocation algorithm of the kth device is formulated as follows:

where f represents the maximum one of the devices with the same pilot as the kth device for the kth device, and the objective of this optimization problem is to let the kth device select the pilot F can be minimized;

To solve the problem of Firstly, establishing an interference diagram with weighted edges among all devices by using an algorithm for minimizing the maximum interference, and if two devices use the same pilot frequency, the weighted value among the devices is ζ _kl in a formula (5); if two devices use different pilot frequencies, the weighting value is 0, and then a pilot frequency distribution mode for minimizing the interference between the largest devices is found by iterating the condition of adopting each pilot frequency on each device until the pilot frequency matrix phi is not changed any more;

The step 3 specifically includes:

Step 301: active detection of the kth device can be formulated as a maximum likelihood problem;

The active user detection problem is essentially a maximum likelihood problem, when the active state of the kth device, a _k, is determined by Pilot signals received by all APsCalculating to obtain an activity state detection value of the kth deviceIf it isThe active user detects correctly, if not, the active user detects the error; the specific calculation process is as follows:

Since all large-scale fades are assumed to be known, then The likelihood function of (2) is expressed as:

In the method, in the process of the invention, When the value denoted as a _k is determinedIs a function of the probability distribution of (1),Representation pair matrixThe sum of its diagonal elements is taken,Representing the base as natural constant e and the index asI pi Q _m represents a value of determinant which is obtained by multiplying the circumferential rate pi by Q _m to form a new matrix;

Wherein the active indications of devices other than the kth device are ignored and considered constant, then the maximum value of equation (10) is equivalent to The active user detection problem for the kth device is therefore expressed as:

I.e. under the constraint of a _k E {0,1}, find the cause The minimum valueA value of D _ρ、Φ、a_k;

step 302: a partial update coordinate descent method is employed to simplify the scheme for detecting device activity,

According to the minimum-maximum covariance interference pilot allocation algorithm, each device is only actively detected by its main AP, and the pilot matrix phi adopted by each device is also selected; for the kth device, its generalized active state identifier Fu _k＝a_kρ_k is defined, and the generalized active state identifier detection value obtained when the active user detection algorithm calculates is recorded asThenThe active user detection problem of (1) is reduced to:

I.e. at Under the constraint of this condition, find the causeThe minimum valueIs a value of (2); solving by coordinate descentIs to (1) optimize the problem ofFirst initialize to matrix σ ² I, where I represents the identity matrix, and use the current one in each iterationSolving forObtainingAnd then solving to obtainUpdatingThe iterative process is explained in detail below;

setting the maximum iteration number as T, taking T as any integer in the set [1, T ], and detecting the generalized active target state identifier detection value of the kth device calculated in the T-1 th iteration process Is marked asDetecting the calculated generalized active mark state identifier in the t-th iteration processIs marked asAnd iterate out t-1Obtained by adding the optimal updated change value d _t calculated in the t-th iteration processI.e.Therefore, the best updated change value d _t of each iteration needs to be found so that the iteration can be performedApproaching the actual value more and more; the solution process of d _t is as follows:

firstly, using Sherman-Morrison updating algorithm, The updating is as follows:

Wherein the method comprises the steps of Is a computable number such thatThe optimization problem of (c) is rewritten as:

I.e. at Under the constraint of this condition, find the minimum value of c _k+log e_k-b_k Is a value of (2); wherein the method comprises the steps ofIs a constant independent of d _t,Is a number associated with d _t;

Thus deriving log e _k-b_k and finding a value that lets it equal 0, the best update change value d _t in the t-th iteration is solved for

To ensure thatAccording to the non-negativity of (2)Calculated outIs a negative number, and a heavy guarantee is added to d _t: I.e. if D _t takesIs then passed throughObtaining the calculated in the t-th iteration processThus, the t-th iteration ends; then starting the t+1st iteration until the maximum iteration number T is reached,

According to the best obtained after the end of T iterationsActive state detection value for kth deviceObtained by (15):

Wherein the method comprises the steps of Is the threshold for active user detection of the kth device whenWhen the active state detection value for the kth deviceTake a value of 1, whenIn the time-course of which the first and second contact surfaces,The value is 0; if the active state detection value of the kth deviceEqual to the active identifier a _k of the kth device, the active user detects correct and if not, detects errors.