Adaptive pilot transmission method for FDD massive MIMO system

Abstract

The application discloses a kind of adaptive pilot transmission methods suitable for FDD Massive MIMO system, utilize the stage stationary nature of channel and divide time-varying non-stationary channel into multiple stationary channel blocks, and respectively carry out coherence time estimation to multiple stationary channel blocks, and the pilot transmission interval currently used is decided by coherence time.The steps mainly include channel correlation estimation and pilot interval selection.The application carries out high confidence interval estimation to channel correlation, controls the error probability of coherence time estimation, and tries to avoid the influence of system throughput caused by overestimation of coherence time;The application estimates coherence time using threshold method, and according to the relationship between threshold and system effective signal-to-noise ratio, the threshold setting method is given, so that the adaptive pilot method can meet the demand of channel capacity for specific task.

Description

TECHNICAL FIELD

[0001] The application relates to a pilot sending method, in particular to an adaptive pilot sending method suitable for an FDD Massive MIMO system. BACKGROUND

[0002] With the development of 5G mobile networks, the demand of users for wireless network throughput is growing exponentially. Massive MIMO is considered as one of the key technologies of 5G networks due to its high spectral efficiency, high energy efficiency and other characteristics. Massive MIMO is equipped with a large-scale antenna array at the base station, which allows multiple users to multiplex the same time-frequency resources, significantly improving the system throughput. At the same time, the large-scale antenna array at the base station can also reduce the loss of channel capacity caused by small-scale fading. However, these advantages are based on the accurate acquisition of channel state information (CSI) at the base station. Since the uplink and downlink channels of the FDD Massive MIMO system work in different frequency bands, there is no channel reciprocity, so the FDD Massive MIMO system needs to estimate the uplink and downlink channel CSI separately. The number of pilots for downlink channel estimation is proportional to the number of base station antennas. When the number of base station antennas is large, the dense pilots significantly reduce the throughput of the communication system. Therefore, to improve the throughput of the communication system, optimizing the sending interval of the pilot is an effective method. By using different pilot sending intervals for channels with different coherence times, the communication resources occupied by the pilot can be reduced, and the throughput of the system can be improved.

[0003] The coherence time of the channel refers to the maximum time length during which the channel is approximately constant, which is related to the user's moving speed. In general, the channel of a low-speed user has a longer coherence time, and the channel of a high-speed user has a shorter coherence time. When the pilot sending interval is less than or equal to the coherence time of the channel, the pilot signal can be used as an accurate estimate of the channel state within the coherence time. When the pilot sending interval is greater than the coherence time of the channel, the channel has changed, and the pilot cannot accurately reflect the current channel state, the channel estimation error increases, and the performance of the entire system is affected. Therefore, the sending interval of the pilot can be set according to the coherence time of the channel, so as to maximize the throughput of the communication system.

[0004] The current pilot adaptive transmission method is mainly divided into two categories, the first method assumes that the wireless propagation environment satisfies the Jakes uniform scattering model, at this time the autocorrelation function of the channel can be expressed as the 0 order Bessel function. Under the condition of knowing the user's moving speed, the first zero point of the 0 order Bessel function corresponds to the time interval which can be used as a rough estimate of the coherence time, and the pilot transmission interval is adjusted according to the estimated value. The second method discretizes the pilot transmission interval into an integer multiple of the minimum transmission interval, and uses the above analytical autocorrelation function to model the pilot interval optimization problem as a combinatorial optimization problem. By using Branch and bound, Tabu Search, Simulated Annealing and other algorithms, the pilot transmission interval that maximizes the system throughput is obtained. However, both of these two methods rely on the assumption of uniform scattering environment, in the case of indoor and non-uniform distribution of scatterers, there is no analytical expression for the autocorrelation function of the channel, at this time it is unreasonable to continue to use the 0 order Bessel function to estimate the coherence time. In addition, both of these two methods assume that the base station knows the current moving speed of the user, but the real-time speed information of the user needs to occupy additional time-frequency resources, which inevitably reduces the throughput of the system. SUMMARY

[0005] In view of the above problems existing in the prior art, the present application provides an adaptive pilot transmission method suitable for FDD Massive MIMO system.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme:

[0007] The adaptive pilot transmission method suitable for FDD Massive MIMO system uses the stage stationary property of the channel to decompose the time-varying non-stationary channel into multiple stationary channel blocks, and estimates the coherence time of each stationary channel block. The coherence time determines the current pilot transmission interval; the method comprises the following steps:

[0008] Step 1: channel correlation estimation; first, the channel is divided into multiple stationary channel blocks, then the correlation of the channel is estimated in the previous predetermined time interval of each stationary channel block by using the independent and identically distributed channel samples sampled on the array antenna;

[0009] Step 2: pilot interval selection; the threshold method is used to give the estimated value of the coherence time, and the pilot interval is set according to the estimated value.

[0010] Further, step 1 comprises:

[0011] Step 1-1: obtain the sample correlation coefficient of the channel by channel samples;

[0012] Step 1-2: Use Fisher-Z transform to give the time correlation coefficient. The high confidence interval.

[0013] Furthermore, step 1-1 specifically includes:

[0014] Let user k correspond to the current time slot of the channel be n, and assume the channel vector g after the m-th time slot. k (n+l) and g k (n) Within the same stable block; let the length of the stable channel block be N. s There are N time slots, and the time window length for correlation coefficient estimation is N. m N time slots; N before the start of each smooth block m In time slot n, the base station obtains the channel sample vector of user k's antenna array in time slot n. This leads to the time correlation coefficient between channel estimates. Under the assumption of independent and identically distributed Rayleigh fading, the time correlation coefficient Based on the sample correlation coefficient r k,l To make an estimate, that is Then, the sample correlation coefficient vectors corresponding to different time intervals l are obtained.

[0015] Furthermore, steps 1-2 specifically include:

[0016] For the correlation coefficient r k,l Perform Fisher-Z transform:

[0017]

[0018] Transformation value z k,l It approximately follows the mean. Standard deviation is The normal distribution; let If the joint confidence level is 1-α, then Corresponding significance level The 1-ζ confidence interval is Where u ζ / 2 It is the upper quantile of ζ / 2 of the standard normal distribution; in obtaining the interval estimate [L] k,l ′,U k,l After '], use the inverse hyperbolic tangent transform to transform [L] k,l ′,U j,l Transform ′] into correlation coefficient 1-ζ confidence interval [L j,l U k,l ]=[tanhL k,l ′,tanhU k,l ′).

[0019] Further, step 2 includes:

[0020] Step 2-1: set the channel correlation threshold b according to the specific system performance requirements;

[0021] Step 2-2: compare the lower bound L of each element in the sample correlation coefficient vector with the set threshold b, and select k,l the element with the maximum value as the interval of the pilot sent in the current stationary block; when the amount of data sent exceeds N s slots, it is considered that the channel enters the next stationary block, and step 1 is repeated. As the interval of the pilot sent in the current stationary block; when the amount of data sent exceeds N s slots, it is considered that the channel enters the next stationary block, and step 1 is repeated.

[0022] Further, in step 2-1, when the downlink signal-to-noise ratio p dl =c, the channel total capacity index of the traditional method is C, and the adaptive algorithm is required to be greater than the channel total capacity index C when p dl ≥d>c; the channel correlation threshold b is set as:

[0023]

[0024] Wherein, wherein β k represents the large-scale fading coefficient between the base station and the kth user; τ dl is the pilot sequence length.

[0025] Further, the channel correlation is estimated by fisher-z transformation with high confidence interval, which is used to control the error probability of coherence time estimation.

[0026] Further, by the threshold setting method based on the effective signal-to-interference-and-noise ratio of the system, the adaptive pilot method can meet the channel total capacity index under high confidence conditions.

[0027] Compared with the prior art, the beneficial effects of the present application are:

[0028] 1. By estimating the channel correlation with high confidence interval, the error probability of coherence time estimation is controlled, and the influence of overestimation of coherence time on system throughput is avoided as much as possible.

[0029] 2. The threshold method is used to estimate the coherence time, and according to the relationship between the threshold and the effective signal-to-interference-and-noise ratio of the system, the threshold setting method is given, so that the adaptive pilot method can meet the demand of channel capacity for specific tasks. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 is the pilot reduction ratio corresponding to different signal-to-noise ratios and user speeds;

[0031] Figure 2is the bit error rate curve when the user speed is 1 m / s;

[0032] Figure 3 is the bit error rate curve when the user speed is 10 m / s;

[0033] Figure 4 is a quasi-static block fading channel model diagram;

[0034] Figure 5 is a schematic diagram of an adaptive pilot transmission method;

[0035] Figure 6 is a flowchart of the method of the present application. DETAILED DESCRIPTION

[0036] The present application will be further described in detail below with reference to the accompanying drawings.

[0037] The method proposed by the present application is based on the following facts: (1) When the antenna distance exceeds half a wavelength, the channels corresponding to different antennas at the same time are approximately independently identically distributed. (2) Channel time correlation describes the similarity of channels at different times, and thus can be used to estimate the correlation time of the channel. (3) The time correlation of the channel changes with the change of the user's moving speed, and the magnitude of the change depends on the user's acceleration. Although the base station cannot obtain the acceleration information, the maximum acceleration of a moving user is generally not greater than one gravitational acceleration, so within a range of several hundred milliseconds, the channel can be considered to be generally stationary. In summary, the pilot interval optimization problem can be decoupled into pilot interval optimization problems within several short time ranges, thereby realizing efficient and scalable adaptive pilot transmission. Based on the above idea, the present application proposes an adaptive pilot transmission method suitable for FDD Massive MIMO systems. This method utilizes the phase stationary property of the channel to decompose the time-varying non-stationary channel into multiple stationary channel blocks, and estimates the coherence time of each stationary channel block. The coherence time determines the currently used pilot transmission interval. In addition, the present application uses high-confidence interval estimation of the coherence time, effectively controlling the error probability of the coherence time estimation, to ensure that the total capacity of the channel of the system meets the service demand.

[0038] As shown in Figure 6 , the present application is an adaptive pilot transmission method suitable for FDD Massive MIMO systems. It includes two steps: channel correlation estimation and pilot interval selection.

[0039] In the channel correlation estimation stage, the channel is first divided into multiple stationary channel blocks, and then the correlation of the channel is estimated in the preceding short time interval of each stationary channel block using the independently and identically distributed channel samples sampled on the array antenna.

[0040] In the pilot interval selection stage, a threshold method is used to give an estimated value of the coherence time, and the pilot interval is set according to the estimated value. Since the actual communication channel is still time-varying within the coherence time, the increase of the pilot interval inevitably causes the increase of the precoding time delay error, and therefore the application provides a threshold design method which can be self-defined according to the system performance requirement. This makes the adaptive pilot method further reduce the pilot overhead and improve the system throughput under the condition of meeting the system performance index.

[0041] The system model of the application is as follows: considering a single-cell FDD Massive MIMO system, the base station has M antennas and simultaneously serves K single-antenna users, M>>K. A quasi-static block fading channel model is adopted, as shown in the formula (1). Figure 4 The channel is approximately constant within a time slot (0.5 ms by default), but varies between different time slots, and the channel has time correlation between time slots. The channel vector between the base station and the kth user is denoted as:

[0042]

[0043] Wherein β k represents the large-scale fading coefficient between the base station and the kth user, represents the influence of small-scale fading. Assuming that the channel satisfies the generalized stationarity within a stationary block, β k remains unchanged within a stationary block. Let the current time slot of the user k corresponding to the channel be n, and assume that the channel vector g k (n+1) and g k (n) are within the same stationary block, then g k (n+1) can be written as

[0044] g k (n+1) = ρ k,l g k (n) + ε k (n+1),

[0045] Wherein is a noise vector independent of g k (n). ρ k,l is the channel time correlation coefficient. In the pilot transmission stage, the base station transmits a downlink pilot matrix to the K users: ψ m represents the pilot sequence transmitted on the mth antenna. Let the pilot sequence length be τ dl , then τ dl ≥ M, Ψ H ψ = I M , I M ​It is an M×M dimensional identity matrix. The user's received signal matrix T p It can be represented as in p dl It is the signal-to-noise ratio of downlink transmission, W p It is an additive unit complex Gaussian noise matrix. User k receives signal Y... p Then, the channel vector g can be obtained. k Estimate The user then sends the CSI directly to the base station in an unquantized and uncoded manner. The base station uses the CSI for linear precoding and transmits the user data.

[0046] Step 1: Channel correlation estimation.

[0047] Sub-step 1: Obtain the sample correlation coefficient of the channel using channel samples. For example... Figure 5 As shown, let the length of the stable block be N. s There are N time slots, and the time window length for correlation coefficient estimation is N. m N time slots. The Nth time slot before the start of each smooth block. m In one time slot, the base station can obtain the channel sample vector of user k's antenna array in time slot n. Since the channel samples are derived from channel estimation, only the time correlation coefficient between channel estimates can be obtained. Under the assumption of independent and identically distributed Rayleigh fading, the correlation coefficient It can be derived from the sample correlation coefficient r k,l Make an estimate, Then, the sample correlation coefficient vectors corresponding to different time intervals l can be obtained.

[0048] Sub-step 2: Use Fisher-Z transform to obtain the correlation coefficient. The high confidence interval. For the correlation coefficient r... k,l Perform Fisher-Z transform:

[0049]

[0050] Transformation value z k,l It approximately follows the mean. Standard deviation is It follows a normal distribution. Let If the joint confidence level is 1-α, then Corresponding significance level The 1-ζ confidence interval is Where u ζ / 2 It is the upper quantile of ζ / 2 of the standard normal distribution. In obtaining the interval estimate [L...] k,l ′,U k,lAfter that, the inverse hyperbolic tangent transformation is used to transform [L k,l , U k,l ] into correlation coefficients 1-ζ confidence interval of [L k,l , U k,l ] = [tanhL k,l , tanhU k,l ].

[0051] Step two: pilot interval selection.

[0052] Sub-step one: set the channel correlation threshold b according to the specific system performance requirements. Since the correlation coefficient of the channel estimation value is , the correlation between the channel estimation values and is as follows

[0053]

[0054] Compared with the traditional pilot transmission method, the channel estimation value can be obtained at each time slot. The base station can only use the last channel estimation in the time slot without transmitting the pilot. After the base station transmits the zero-forcing precoded signal x, the effective signal-to-interference-and-noise ratio (SINR) of the received signal at user k under the confidence level of 1-α is adapt,k .

[0055]

[0056] The effective signal-to-interference-and-noise ratio (SINR) of user k under the traditional method is conv,k .

[0057]

[0058] Let the downlink signal-to-noise ratio p dl = c, and the total channel capacity index of the traditional method is C. The goal of setting the threshold is to require the adaptive algorithm to be greater than the total channel capacity index C when p dl ≥ d > c. Since SINR adapt,k is monotonically increasing with respect to p dl , and the total channel capacity is monotonically increasing with respect to SINR adapt,k . Therefore, when , the total channel capacity of the adaptive method can be guaranteed to be greater than C with a confidence level of 1-α when p dl ≥ d > c. At this time, the threshold b is set to

[0059]

[0060] Sub-step two: transform the sample correlation coefficient vector The lower bound L of each element k,l Compare with the set threshold b, and select This serves as the interval for pilot transmission within the current stable block. When the amount of data transmitted exceeds N... s If a time slot is reached, it is considered that the channel has entered the next stable block, and step one is repeated.

[0061] Experimental results show that the adaptive pilot method proposed in this invention performs well under high signal-to-noise ratio and low-speed conditions, reducing pilot frequency by approximately 80%. Figure 1 As shown. Furthermore, compared to existing methods based on the Jakes channel model, this adaptive pilot method can significantly reduce the risk of overestimating channel coherence time, reducing the number of pilots transmitted while maintaining system performance. Figure 2 and Figure 3 These are comparisons of algorithm performance when the Jakes channel model underestimates and overestimates.

[0062] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An adaptive pilot transmission method suitable for FDD Massive MIMO systems, characterized in that, The time-varying non-stationary channel is decomposed into multiple stationary channel blocks by using the phase stationary property of the channel, and the coherent time is estimated for the multiple stationary channel blocks respectively, and the pilot transmission interval used currently is determined by the coherent time; the method comprises the following steps: Step 1: channel correlation estimation; first, the channel is divided into multiple stationary channel blocks, then the correlation of the channel is estimated by using the independent and identically distributed channel samples sampled on the array antenna in a predetermined time period before each stationary channel block; step 1 comprises: Step 1-1: the sample correlation coefficient of the channel is obtained by using the channel samples; step 1-1 specifically comprises: Allow users The corresponding channel current timeslot is Assuming the first Channel vector after one time slot and Within the same stable block; let the length of the stable channel block be... There are 1 time slot, and the time window length for correlation coefficient estimation is 1. One time slot; before the start of each smooth block In each time slot, the base station obtains user information. In the time slot Channel sample vector of upper antenna array This leads to the time correlation coefficient between channel estimates. Under the assumption of independent and identically distributed Rayleigh fading, the time correlation coefficient From the sample correlation coefficient To make an estimate, that is Then, different time intervals were obtained. The corresponding sample correlation coefficient vector [ ]; Steps 1-2: Utilizing transform to give time-dependent coefficients of high confidence interval; Step 1-2 specifically comprises: On the correlation coefficient The Transform: ； transformed values approximately follow a normal distribution with mean and standard deviation ; let be the joint confidence level for , then the corresponding significance level , the interval for is , where is the upper quantile of the standard normal distribution; after obtaining the interval estimate , transform to the interval for the channel correlation coefficient ; Step 2: pilot interval selection; the estimated value of the coherent time is given by using the threshold method, and the pilot interval is set according to the estimated value; step 2 comprises: Step 2-1: Set channel correlation threshold according to specific system performance requirement ; Step 2-2: Convert the sample correlation coefficient vector [ The lower bound of each element in ] With the set threshold Compare and select This serves as the interval for pilot transmission within the current stable block; when the amount of data transmitted exceeds... If a time slot is reached, it is considered that the channel has entered the next stable block, and step 2-1 is repeated. 2.The adaptive pilot transmission method for FDD Massive MIMO system of claim 1, wherein, In step 2-1, let the downlink signal-to-noise ratio be The total channel capacity index of the traditional method is , the adaptive algorithm is required to be greater than the total channel capacity index when ; and the channel correlation threshold is set as ； wherein, wherein represents a large-scale fading coefficient between the base station and the user; is a pilot sequence length. 3.The adaptive pilot transmission method for FDD Massive MIMO system of claim 1, wherein, By Transforming the correlation of the channel makes a high-confidence interval estimate, which is used to control the probability of error in the coherence time estimate. 4.The adaptive pilot transmission method for FDD Massive MIMO system of claim 2, wherein, By using the threshold setting method based on the system effective signal-to-interference-and-noise ratio, it is ensured that the adaptive pilot method meets the channel total capacity index under the condition of high confidence.

Images