CN105634709A - Pilot frequency distribution method - Google Patents

Abstract

The invention provides a pilot frequency distribution method. The method can be used for effectively solving the problem of serious pilot frequency pollution between adjacent cell users in a large-scale MIMO system in actual network architecture and a transmission environment. The method comprises the following steps: a step 1, generating a network level base sequence; a step 2, generating a cell level base sequence; a step 3, generating a user level base sequence; a step 4, calculating statistical mutual correlation values between user pilot frequency sequences and the pilot frequency sequences of the adjacent cell; a step 5, sequencing and grouping the user pilot frequency sequences in each cell according to the statistical mutual correlation values; a step 6, determining a pilot frequency group to which a user belongs; a step 7, allocating the user pilot frequency; a step 8, updating channel estimation; and a step 9, updating the user pilot frequency. The method can be used for effectively reducing the serious pilot frequency pollution between adjacent cell users resulting from a limited number of orthogonal pilot sequences in the MIMO system, and improving the actual performance of the MIMO technology; and the method is simple to implement, relatively low in computational complexity and is relatively good in realizability.

Description

A kind of pilot distribution method

Technical field

The present invention relates to a kind of pilot distribution method, belong to communication technology application.

Background technology

The MIMO technology having obtained practical application at present is generally based on fewer number of aerial array (being typically less than 10 array elements). In recent years, a kind of novel large scale array MIMO (MassiveMIMO) technology gets the attention. Its main array number amount being characterised by aerial array is significantly increased, and commonly reaches dozens or even hundreds of array element. Communication system based on this antenna can realize significantly improving of system transfers reliability and the rate of information throughput by simple signal processing. Namely assume when base station has imperfect channel state information (now each user is pairwise orthogonal to the channel of base station), by simple linear process (such as, maximum-ratio combing) or ZF process, just can eliminate the interference of other users under without more frequency spectrum resource. Such process drastically increases spectrum efficiency. Additionally, in base station by using a very big antenna array, it is possible to greatly reduce the energy that system transfers consumes.

But, in practical application, base station does not often have imperfect channel state information condition, therefore, traditional method based on pilot training sequence generally still can be adopted to carry out channel estimating to obtain the time of day of channel. In the multiple cell scene that cellular system is common, also it is limited owing to frequency spectrum resource distribution and the channel coherency time of pilot transmission are all the quantity of limited, orthogonal pilot frequency sequence. Therefore, pilot frequency sequence must reuse in the user of other communities. So, multipath fading (multipath fading) channel estimating in given community will be polluted by the pilot transmission in other communities, thus causing serious pilot pollution, have impact on the actual performance of the precision of multipath fading channel estimating and system. Therefore, a lot of research worker propose various channel estimation methods to reduce the impact of pilot pollution, improve the accuracy of channel estimating. Wherein important a kind of resolving ideas is around design and the distribution of pilot tone.

One of which method is by utilizing the orthogonality of channel vector in extensive mimo system, it is proposed that a kind of pilot frequency sequence based on Chu sequence generates and distribution method. The method utilizes the control parameter that design is different, it is possible to make in community in the completely orthogonal situation of the pilot frequency sequence of each user, reduces the pilot pollution problem between inter-cell user simultaneously as far as possible, makes pilot pollution obtain a degree of alleviation.

Existing pilot tone generates and in distribution method, mainly considers the control parameter being chosen different districts by the criterion designed, so that have good cross correlation between the cell-level pilot tone basic sequence of different districts, thus reducing pilot pollution.

But the pilot frequency sequence that this method is determined is fixing, and have employed the distribution method of static state.

It is impossible to ensure that in completely orthogonal situation between user's pilot frequency sequence between neighbor cell, between two pilot tone between cross correlation be not constant, certainly exist the cross correlation between some pilot frequency sequence and other pilot tones better, and some situation poor all the time, therefore, can consider to utilize this feature, good for cross correlation pilot tone is distributed to the user being in cell edge, and the poor pilot tone of cross correlation distributes to the user being in center of housing estate, thus realizing the dynamic distribution of pilot tone.

In addition, consider from the angle of interference randomization, pilot tone for user is distributed, it is contemplated that and carries out regular renewal, by periodically distributing different pilot frequency sequences, so that the pilot pollution between user is on close level unanimously, thus avoiding have be in more weak pilot pollution all the time owing to the uneven certain user of making of pilot tone characteristic is subject to more serious pilot pollution all the time, the fairness of pilot pollution level or even service quality between user can be improved.

Therefore, it is necessary on the basis of existing pilot frequency sequence static allocation, it is considered to dynamic allocation scheme.

Summary of the invention

The pilot pollution problem serious between the neighbor cell user that can effectively solve under real network framework and communication environments in extensive mimo system.

(1) the user class pilot frequency sequence packet of Corpus--based Method cross-correlation;

(2) the ownership pilot packet of user is determined based on large scale decline;

(3) distribution in groups in network topology it is grouped in based on the user class pilot frequency sequence of soft-frequency reuse thought;

(4) update based on user's pilot frequency sequence of interference randomization thought.

A kind of pilot distribution method, this method comprises the steps,

Step 1: generate network level basic sequence

Generation method according to Chu sequence, generates network level pilot tone basic sequence a=[a₀, a₁..., a_��-1], �� is sequence length. Wherein a_nProduced by equation below:

${\mathrm{\α}}_n=\exp i\frac{N\mathrm{\π}}{\mathrm{\τ}}n(n+(\mathrm{\τ}\mathrm{mod}2\left)\right)$

N=0,1 ..., ��-1 (1)

Wherein i is imaginary unit, and N is for controlling parameter, and it should be the integer relatively prime with ��; a₀, a₁..., a_��-1For being exactly every element of sequence a, whole sequence is pilot frequency sequence.

Step 2: generate cell-level basic sequence

According to the multiplexing factor in the topological structure of cellular cell in network design, set and need the cell-level basic sequence number that generates as L. Considering that cell-level basic sequence can be reused in the community outside the first interference circle, general L is taken as 3 or 7.

To jth community, on the basis of network level basic sequence, the skew of enterprising line phase, generates the cell-level basic sequence of this community�� _j, n-th of this sequence calculates according to formula below:

${\[{\underset{\‾}{\mathrm{\ψ}}}_j\]}_n=a_n\exp i\frac{2{\mathrm{\πq}}_{j}n}{\mathrm{\τ}}$

N=1,2 ... �� (2)

Wherein q_jPhase offset for this community controls parameter, and it can not be the integral multiple of ��, and can not be equal with j.

Step 3: generate user class pilot frequency sequence

On the basis of cell-level pilot tone basic sequence, obtained the user class pilot frequency sequence of this community by cyclic shift. Namely assume that single subdistrict receivability number of users is K, then jth community kth user's pilot frequency sequence (k=1,2 ..., K) be:

�� _jk=<�� _j��_k-1(3)

Wherein < >_kRepresent vector cyclic shift k position to the left.

Step 4: calculate the statistics cross correlation value between each pilot frequency sequence of each user's pilot frequency sequence and adjacent area

First maximum offset S when calculating pilot frequency sequence cross-correlation between neighbor cell is determined. The big I of S is according to MPS process scene configuration, namely S is equal to the time domain length divided by code element single in pilot frequency sequence of the average propagation time delay from base station to cell edge under this covering scene, it is possible to it is set to 0 (namely only considering 0 value cross-correlation) for the purpose of simplification.

To L the community constructed �� K user=LK user class pilot frequency sequence, calculate outside unique user pilot tone and this community the statistical average of cross correlation value between all user's pilot tones one by one respectively, that is, the kth user's pilot tone for jth community, its statistics cross correlation value is:

${\mathrm{SC}}_{jk}=\frac{1}{(L-1)\&CenterDot;K\&CenterDot;S}\underset{m=1,m\&NotEqual;j}{\overset{L}{\&Sigma;}}\underset{n=1}{\overset{K}{\&Sigma;}}\underset{s=0}{\overset{S}{\&Sigma;}}{\mathrm{ccor}}_s({\underset{\&OverBar;}{\mathrm{\&psi;}}}_{jk},{\underset{\&OverBar;}{\mathrm{\&psi;}}}_{mn})$

Wherein ccor_s(a b) represents that between sequence a and b, side-play amount is the cross correlation value of s.

Step 5: the user's pilot frequency sequence in each community is ranked up according to its statistics cross correlation value and is grouped

K the user's sequence setting each community can be divided into 2 groups, is central. set and border group respectively, and the ratio that often in group, the number of sequence accounts for whole community user sequence total can configure as required, is defaulted as 1/2, namely K/2.

Then K user's sequence to each community, is ranked up from big to small according to the statistics cross correlation value calculated in step 4, and K/2 user's pilot tone above is attributed to central. set, and K/2 below is attributed to border group.

Step 1��5 are precalculated. Step 6��9 are the user's pilot tone assigning process in real network.

Step 6: determine the pilot packet belonging to user

Current large scale fading characteristic according to actual user each in network determines the pilot packet of user.

Central. set: when the meansigma methods of a user to the large scale decline of all neighbor base stations having neighboring BS relationship with it is higher than setting thresholding, or its dispersion declined to the large scale of each neighbor base station is less than certain thresholding, then group centered by this user's current home.

Border group: when a user has the dispersion of the large scale decline of all neighbor base stations of neighboring BS relationship more than certain thresholding to it, or when it declines more than certain thresholding to large scale of certain neighbor base station, then this user's current home is border group.

Step 7: user's pilot tone is distributed

Affiliated packet according to this user, randomly chooses a unallocated pilot tone from corresponding user's sequence of packets of this user affiliated subdistrict and distributes to this user.

Step 8: channel estimating updates

This distribution keeps M frame constant (updating Tong Bu with large scale decline), every frame carries out channel estimating (single community MMSE or multiple cell MMSE or other linear channel algorithm for estimating), updates each user and decline to the large scale of each adjacent base station after every M frame.

Step 9: user's pilot updating

According to the large scale decline after updating, the whole network user is redefined its affiliated pilot packet, and in group random assortment pilot tone. Even if there is not any change in this user affiliated subdistrict and affiliated pilot packet, it is also necessary to randomly choose a unallocated pilot tone from corresponding user's sequence of packets of this user affiliated subdistrict and distribute to this user (based on interference randomization thought).

According to the pilot frequency sequence that above method generates, there is following characteristic, all meet orthogonal between the multiple cell-level basic sequences namely produced by a network level basic sequence; And be also orthogonal between all user's sequences produced by a cell-level basic sequence, and it cannot be guaranteed that orthogonal between user's sequence of different districts.

If large scale (path loss and shade) decline is it is known that be left out updating signaling consumption when pilot tone is distributed.

Compared with prior art, the present invention has the advantages that.

1, can effectively reduce pilot pollution serious between the neighbor cell user caused due to orthogonal pilot frequency sequence limited amount in mimo system, improve the actual performance of MIMO technology.

2, this method realizes simplicity, and computational complexity is relatively low, has good realizability.

Accompanying drawing explanation

Fig. 1 is the implementing procedure figure of this method.

Detailed description of the invention

Being illustrated in figure 1 the implementation flow process of the present invention, this method comprises the steps,

Step 1: generate network level basic sequence

Generation method according to Chu sequence, generates network level pilot tone basic sequence a=[a₀, a₁..., a_��-1], �� is sequence length. Wherein a_nProduced by equation below:

$a_n=\exp i\frac{N\mathrm{\&pi;}}{\mathrm{\&tau;}}N(n+(\mathrm{\&tau;}\mathrm{mod}2\left)\right)$

N=0,1 ..., ��-1 (1)

Wherein i is imaginary unit, and N is for controlling parameter, and it should be the integer relatively prime with ��; a₀, a₁..., a_��-1For being exactly every element of sequence a, whole sequence is pilot frequency sequence

Step 2: generate cell-level basic sequence

According to the multiplexing factor in the topological structure of cellular cell in network design, set and need the cell-level basic sequence number that generates as L. Considering that cell-level basic sequence can be reused in the community outside the first interference circle, general L is taken as 3 or 7.

To jth community, on the basis of network level basic sequence, the skew of enterprising line phase, generates the cell-level basic sequence of this community�� _j, n-th of this sequence calculates according to formula below:

${\&lsqb;{\underset{\&OverBar;}{\mathrm{\&psi;}}}_j\&rsqb;}_n=a_n\exp i\frac{2{\mathrm{\&pi;q}}_{j}n}{\mathrm{\&tau;}}$

N=1,2 ... �� (2)

Wherein q_jPhase offset for this community controls parameter, and it can not be the integral multiple of ��, and can not be equal with j.

Step 3: generate user class pilot frequency sequence

On the basis of cell-level pilot tone basic sequence, obtained the user class pilot frequency sequence of this community by cyclic shift. Namely assume that single subdistrict receivability number of users is K, then jth community kth user's pilot frequency sequence (k=1,2 ..., K) be:

�� _jk=<�� _j��_k-1(3)

Wherein < >_kRepresent vector cyclic shift k position to the left.

Step 4: calculate the statistics cross correlation value between each pilot frequency sequence of each user's pilot frequency sequence and adjacent area

First maximum offset S when calculating pilot frequency sequence cross-correlation between neighbor cell is determined. The big I of S is according to MPS process scene configuration, namely S is equal to the time domain length divided by code element single in pilot frequency sequence of the average propagation time delay from base station to cell edge under this covering scene, it is possible to it is set to 0 (namely only considering 0 value cross-correlation) for the purpose of simplification.

To L the community constructed �� K user=LK user class pilot frequency sequence, calculate outside unique user pilot tone and this community the statistical average of cross correlation value between all user's pilot tones one by one respectively, that is, the kth user's pilot tone for jth community, its statistics cross correlation value is:

${\mathrm{SC}}_{jk}=\frac{1}{(L-1)\&CenterDot;K\&CenterDot;S}\underset{m=1,m\&NotEqual;j}{\overset{L}{\&Sigma;}}\underset{n=1}{\overset{K}{\&Sigma;}}\underset{s=0}{\overset{S}{\&Sigma;}}{\mathrm{ccor}}_s({\underset{\&OverBar;}{\mathrm{\&psi;}}}_{jk},{\underset{\&OverBar;}{\mathrm{\&psi;}}}_{mn})-$

Wherein ccor_s(a b) represents that between sequence a and b, side-play amount is the cross correlation value of s.

Step 5: the user's pilot frequency sequence in each community is ranked up according to its statistics cross correlation value and is grouped

K the user's sequence setting each community can be divided into 2 groups, is central. set and border group respectively, and the ratio that often in group, the number of sequence accounts for whole community user sequence total can configure as required, is defaulted as 1/2, namely K/2.

Then K user's sequence to each community, is ranked up from big to small according to the statistics cross correlation value calculated in step 4, and K/2 user's pilot tone above is attributed to central. set, and K/2 below is attributed to border group.

Step 1��5 are precalculated. Step 6��9 are the user's pilot tone assigning process in real network.

Step 6: determine the pilot packet belonging to user

Current large scale fading characteristic according to actual user each in network determines the pilot packet of user.

Central. set: when the meansigma methods of a user to the large scale decline of all neighbor base stations having neighboring BS relationship with it is higher than setting thresholding, or its dispersion declined to the large scale of each neighbor base station is less than certain thresholding, then group centered by this user's current home.

Border group: when a user has the dispersion of the large scale decline of all neighbor base stations of neighboring BS relationship more than certain thresholding to it, or when it declines more than certain thresholding to large scale of certain neighbor base station, then this user's current home is border group.

Step 7: user's pilot tone is distributed

Affiliated packet according to this user, randomly chooses a unallocated pilot tone from corresponding user's sequence of packets of this user affiliated subdistrict and distributes to this user.

Step 8: channel estimating updates

This distribution keeps M frame constant (updating Tong Bu with large scale decline), and every frame carries out channel estimating (list community MMSE or multiple cell MMSE), updates each user and decline to the large scale of each adjacent base station after every M frame.

Step 9: user's pilot updating

According to the large scale decline after updating, the whole network user is redefined its affiliated pilot packet, and in group random assortment pilot tone. Even if there is not any change in this user affiliated subdistrict and affiliated pilot packet, it is also necessary to randomly choose a unallocated pilot tone from corresponding user's sequence of packets of this user affiliated subdistrict and distribute to this user (based on interference randomization thought).

According to the pilot frequency sequence that above method generates, there is following characteristic, all meet orthogonal between the multiple cell-level basic sequences namely produced by a network level basic sequence; And be also orthogonal between all user's sequences produced by a cell-level basic sequence, and it cannot be guaranteed that orthogonal between user's sequence of different districts.

If large scale decline (path loss and shade) is it is known that be left out updating signaling consumption when pilot tone is distributed.

Claims (2)

1. a pilot distribution method, it is characterised in that: this method comprises the steps,

Step 1: generate network level basic sequence

Generation method according to Chu sequence, generates network level pilot tone basic sequence a=[a₀, ��₁..., a_��-1], �� is sequence length; Wherein a_nProduced by equation below:

$a_n=\exp i\frac{N\mathrm{\&pi;}}{\mathrm{\&tau;}}n(n+(\mathrm{\&tau;}\mathrm{mod}2\left)\right)$

N=0,1 ..., ��-1 (1)

Wherein i is imaginary unit, and N is for controlling parameter, and it should be the integer relatively prime with ��; a₀, a₁..., ��_��-1For being exactly every element of sequence a, whole sequence is pilot frequency sequence;

Step 2: generate cell-level basic sequence

According to the multiplexing factor in the topological structure of cellular cell in network design, set and need the cell-level basic sequence number that generates as L; Considering that cell-level basic sequence can be reused in the community outside the first interference circle, general L is taken as 3 or 7;

To jth community, on the basis of network level basic sequence, the skew of enterprising line phase, generates the cell-level basic sequence of this community�� _j, n-th of this sequence calculates according to formula below:

${\&lsqb;{\underset{\&OverBar;}{\mathrm{\&psi;}}}_j\&rsqb;}_n=a_n\exp i\frac{2{\mathrm{\&pi;q}}_{j}n}{\mathrm{\&tau;}}$

N=1,2 ... �� (2)

Wherein q_jPhase offset for this community controls parameter, and it can not be the integral multiple of ��, and can not be equal with j;

Step 3: generate user class pilot frequency sequence

On the basis of cell-level pilot tone basic sequence, obtained the user class pilot frequency sequence of this community by cyclic shift; Namely assume that single subdistrict receivability number of users is K, then jth community kth user's pilot frequency sequence (k=1,2 ..., K) be:

�� _jk=<��j>_k-1(3)

Wherein < >_kRepresent vector cyclic shift k position to the left;

Step 4: calculate the statistics cross correlation value between each pilot frequency sequence of each user's pilot frequency sequence and adjacent area

First maximum offset S when calculating pilot frequency sequence cross-correlation between neighbor cell is determined; The big I of S is according to MPS process scene configuration, and namely S is equal to the time domain length divided by code element single in pilot frequency sequence of the average propagation time delay from base station to cell edge under this covering scene, it is possible to it is set to 0 for the purpose of simplification;

To L the community constructed �� K user=LK user class pilot frequency sequence, calculate outside unique user pilot tone and this community the statistical average of cross correlation value between all user's pilot tones one by one respectively, that is, the kth user's pilot tone for jth community, its statistics cross correlation value is:

${\mathrm{SC}}_{jk}=\frac{1}{(L-1)\&CenterDot;K\&CenterDot;S}\underset{m=1,m\&NotEqual;j}{\overset{L}{\&Sigma;}}\underset{n=1}{\overset{K}{\&Sigma;}}\underset{s=0}{\overset{S}{\&Sigma;}}{\mathrm{ccor}}_s({\underset{\&OverBar;}{\mathrm{\&psi;}}}_{jk},{\underset{\&OverBar;}{\mathrm{\&psi;}}}_{mn})$

Wherein ccor_s(a, b) represents that between sequence a and b, side-play amount is the cross correlation value of s;

Step 5: the user's pilot frequency sequence in each community is ranked up according to its statistics cross correlation value and is grouped

K the user's sequence setting each community can be divided into 2 groups, is central. set and border group respectively, and the ratio that often in group, the number of sequence accounts for whole community user sequence total can configure as required, is defaulted as 1/2, namely K/2;

Then K user's sequence to each community, is ranked up from big to small according to the statistics cross correlation value calculated in step 4, and K/2 user's pilot tone above is attributed to central. set, and K/2 below is attributed to border group;

Step 1��5 are precalculated; Step 6��9 are the user's pilot tone assigning process in real network;

Step 6: determine the pilot packet belonging to user

Current large scale fading characteristic according to actual user each in network determines the pilot packet of user;

Central. set: when the meansigma methods of a user to the large scale decline of all neighbor base stations having neighboring BS relationship with it is higher than setting thresholding, or its dispersion declined to the large scale of each neighbor base station is less than certain thresholding, then group centered by this user's current home;

Border group: when a user has the dispersion of the large scale decline of all neighbor base stations of neighboring BS relationship more than certain thresholding to it, or when it declines more than certain thresholding to large scale of certain neighbor base station, then this user's current home is border group;

Step 7: user's pilot tone is distributed

Affiliated packet according to this user, randomly chooses a unallocated pilot tone from corresponding user's sequence of packets of this user affiliated subdistrict and distributes to this user;

Step 8: channel estimating updates

This distribution keeps M frame constant, and every frame carries out channel estimating, updates each user and decline to the large scale of each adjacent base station after every M frame;

Step 9: user's pilot updating

According to the large scale decline after updating, the whole network user is redefined its affiliated pilot packet, and in group random assortment pilot tone; Even if there is not any change in this user affiliated subdistrict and affiliated pilot packet, it is also necessary to randomly choose a unallocated pilot tone from corresponding user's sequence of packets of this user affiliated subdistrict and distribute to this user;

According to the pilot frequency sequence that above method generates, there is following characteristic, all meet orthogonal between the multiple cell-level basic sequences namely produced by a network level basic sequence; And be also orthogonal between all user's sequences produced by a cell-level basic sequence, and it cannot be guaranteed that orthogonal between user's sequence of different districts.

2. a kind of pilot distribution method according to claim 1, it is characterised in that: if large scale and path loss and shade are it is known that be left out updating signaling consumption when pilot tone is distributed.