CN105634709B - A kind of pilot distribution method - Google Patents

Abstract

A pilot allocation method, which can effectively solve the serious pilot pollution problem between adjacent cell users in a massive MIMO system under the actual network architecture and propagation environment. The method comprises the following steps, step 1: generating a network-level base sequence; step 2: generating a cell-level base sequence; step 3: generating a user-level pilot sequence; step 4: calculating each user pilot sequence and each pilot sequence of a neighboring cell The statistical cross-correlation value between; Step 5: sort and group the user pilot sequences in each cell according to its statistical cross-correlation value; Step 6: Determine the pilot group to which the user belongs; Step 7: Assign user pilots; Step 8: Channel estimation update; Step 9: User pilot update; This method can effectively reduce the severe pilot pollution between users in adjacent cells caused by the limited number of orthogonal pilot sequences in MIMO systems, and improve the reliability of MIMO technology. Actual performance: This method is easy to implement, has low computational complexity, and has good realizability.

Description

A kind of pilot allocation method

technical field

The invention relates to a pilot allocation method, which belongs to the application field of communication technology.

Background technique

The MIMO technology that has been practically applied at present is generally based on a small number of antenna arrays (usually less than 10 array elements). In recent years, a new massive array MIMO (Massive MIMO) technology has received extensive attention. Its most important feature is that the number of array elements of the antenna array has been significantly increased, generally reaching dozens or even hundreds of array elements. The communication system based on this antenna can significantly improve the system transmission reliability and information transmission rate through simple signal processing. That is, assuming that the base station has ideal channel state information (at this time, the channels from each user to the base station are orthogonal), through simple linear processing (for example, maximum ratio combining) or zero-forcing processing, without much The interference of other users can be eliminated under the spectrum resources. Such processing greatly improves spectral efficiency. In addition, by using a very large antenna array at the base station, the energy consumed by system transmission can be greatly reduced.

However, in practical applications, the base station often does not have ideal channel state information conditions, so the traditional pilot training sequence-based method is generally still used for channel estimation to obtain the true state of the channel. In a common multi-cell scenario in a cellular system, since the spectrum resource allocation and channel coherence time for pilot transmission are limited, the number of orthogonal pilot sequences is also limited. Therefore, the pilot sequence must be reused among users in other cells. Therefore, the multipath fading (small-scale fading) channel estimation in a given cell will be polluted by the pilot transmissions in other cells, resulting in serious pilot pollution, which affects the accuracy of small-scale fading channel estimation and the actual performance of the system. performance. Therefore, many researchers have proposed various channel estimation methods to reduce the influence of pilot pollution and improve the accuracy of channel estimation. One of the important solutions is around the design and allocation of pilots.

One of the methods is to propose a Chu-sequence-based pilot sequence generation and allocation method by exploiting the orthogonality of channel vectors in massive MIMO systems. By using different control parameters, this method can make the pilot sequence of each user in the cell completely orthogonal, and at the same time reduce the pilot pollution problem between users in the cell as much as possible, so that the pilot pollution can be reduced to a certain extent. ease.

In the existing pilot generation and allocation methods, the main consideration is to select the control parameters of different cells through the design criteria, so that the cell-level pilot base sequences of different cells have better cross-correlation characteristics, thereby reducing the number of pilots. frequent pollution.

However, the pilot sequence determined by this method is fixed, and a static allocation method is adopted.

In the case that the user pilot sequences between adjacent cells cannot be guaranteed to be completely orthogonal, the cross-correlation characteristics between any two pilot pairs are not constant, and there must be differences between some pilot sequences and other pilot sequences. The cross-correlation characteristics among them are good, but some cases are always poor. Therefore, it can be considered to use this feature to allocate the pilots with better cross-correlation characteristics to the users at the edge of the cell, and the pilots with poor cross-correlation characteristics The frequency is allocated to the user in the center of the cell, so as to realize the dynamic allocation of the pilot.

In addition, from the perspective of interference randomization, regular updates should be considered for the pilot allocation of users. By periodically allocating different pilot sequences, the pilot pollution levels among users can be made close to the same, thereby avoiding the The unevenness of the pilot characteristics makes some users always suffer from relatively serious pilot pollution and some users are always in weak pilot pollution, which improves the fairness of the level of pilot pollution among users and even the quality of service.

Therefore, it is necessary to consider a dynamic allocation strategy on the basis of the existing static allocation of pilot sequences.

Contents of the invention

It can effectively solve the serious pilot pollution problem between adjacent cell users in the massive MIMO system under the actual network architecture and propagation environment.

(1) user-level pilot sequence grouping based on statistical cross-correlation;

(2) Determine the user's home pilot grouping based on large-scale fading;

(3) Group allocation of user-level pilot sequence groups in network topology based on the idea of soft frequency reuse;

(4) User pilot sequence update based on the idea of interference randomization.

A pilot allocation method, the method includes the following steps,

Step 1: Generate network-level base sequences

According to the generation method of the Chu sequence, a network-level pilot base sequence a=[a ₀ , a ₁ ,...,a _τ-1 ] is generated, where τ is the sequence length. where a _n is generated by the following formula:

n=0, 1, ..., τ-1 (1)

Where i is the imaginary number unit, N is the control parameter, which should be an integer relatively prime to τ; a ₀ , a ₁ ,..., a _τ-1 are the elements of the sequence a, and the whole sequence is the pilot sequence.

Step 2: Generate cell-level base sequences

According to the multiplexing factor in the topology structure of the cell in the network deployment, the number of cell-level base sequences to be generated is set as L. Considering that the cell-level base sequence can be reused in cells outside the first interference circle, generally L is taken as 3 or 7.

For the jth cell, the phase offset is performed on the basis of the network-level base sequence to generate the cell-level base sequence ψ _j of the cell, and the nth item of the sequence is calculated according to the following formula:

n=1, 2, . . . τ (2)

Where q _j is the phase offset control parameter of the cell, which cannot be an integer multiple of τ, and cannot be equal to j.

Step 3: Generate user-level pilot sequences

On the basis of the cell-level pilot base sequence, the user-level pilot sequence of the cell is obtained through cyclic shift. That is, assuming that the number of users that can be accommodated in a single cell is K, the kth user pilot sequence (k=1,2,...,K) of the jth cell is:

ψ _jk ＝〈 ψ _j 〉 _k-1 (3)

Among them, 〈·〉 _k means that the vector is cyclically shifted to the left by k bits.

Step 4: Calculate the statistical cross-correlation value between each user pilot sequence and each pilot sequence of the neighboring cell

Firstly, determine the maximum offset S when calculating the cross-correlation of pilot sequences between adjacent cells. The size of S can be configured according to the cell coverage scenario, that is, S is equal to the average propagation delay from the base station to the edge of the cell in the coverage scenario divided by the time domain length of a single symbol in the pilot sequence, or it can be set to 0 for simplicity (i.e. only 0-valued cross-correlations are considered).

For the constructed L cells × K users = LK user-level pilot sequences, calculate the statistical average of the cross-correlation values between the single user pilot and all user pilots outside the cell, that is, for the first For the kth user pilot of cell j, its statistical cross-correlation value is:

Where ccor _s (a, b) represents the cross-correlation value between sequence a and b with offset s.

Step 5: Sort and group the user pilot sequences in each cell according to their statistical cross-correlation values

It is set that the K user sequences in each cell can be divided into 2 groups, which are the center group and the boundary group. The ratio of the number of sequences in each group to the total number of user sequences in the entire cell can be configured according to needs, and the default is 1/2 , namely K/2 pieces.

Then, sort the K user sequences in each cell according to the statistical cross-correlation values calculated in step 4 from large to small, the front K/2 user pilots belong to the central group, and the following K/2 user pilots belong to for the boundary group.

Steps 1 to 5 are pre-calculated. Steps 6-9 are the process of user pilot allocation in the actual network.

Step 6: Determine the pilot group to which the user belongs

The user's pilot group is determined according to the current large-scale fading characteristics of each actual user in the network.

Central group: when the average value of large-scale fading from a user to all neighboring base stations that have a relationship with it is higher than the set threshold, or the dispersion of large-scale fading to each neighboring base station is smaller than a certain threshold, then the The user currently belongs to the central group.

Boundary group: when the dispersion of large-scale fading from a user to all neighboring base stations with which it has a neighboring cell relationship is greater than a certain threshold, or when the large-scale fading to a certain neighboring base station is greater than a certain threshold, the user currently belongs to for the boundary group.

Step 7: User Pilot Allocation

According to the group to which the user belongs, an unassigned pilot is randomly selected from the corresponding user sequence group of the cell to which the user belongs and assigned to the user.

Step 8: Channel estimate update

The allocation remains unchanged for M frames (synchronized with large-scale fading updates), and channel estimation is performed every frame (single-cell MMSE or multi-cell MMSE or other linear channel estimation algorithms), and each user is updated to each neighboring base station after every M frame large-scale decline.

Step 9: User Pilot Update

According to the updated large-scale fading, the pilot groups to which users in the whole network belong are re-determined, and the pilots are randomly assigned in the group. Even if there is no change in the cell to which the user belongs and the pilot group to which it belongs, an unassigned pilot must be randomly selected from the corresponding user sequence group in the cell to which the user belongs (based on the idea of interference randomization).

The pilot sequence generated according to the above method has the following characteristics, that is, the plurality of cell-level base sequences generated by a network-level base sequence are orthogonal; and all user sequences generated by a cell-level base sequence are orthogonal It is also orthogonal, but the user sequences of different cells cannot be guaranteed to be orthogonal.

If the large-scale (path loss and shadow) fading is known, the signaling overhead when updating the pilot allocation is not considered.

Compared with the prior art, the present invention has the following beneficial effects.

1. It can effectively reduce the severe pilot pollution between users in adjacent cells caused by the limited number of orthogonal pilot sequences in the MIMO system, and improve the actual performance of the MIMO technology.

2. The method is easy to implement, has low computational complexity, and has good realizability.

Description of drawings

Fig. 1 is the implementation flowchart of this method.

Detailed ways

As shown in Figure 1, it is the implementation method process flow of the present invention, and this method comprises the following steps,

Step 1: Generate network-level base sequences

According to the generation method of the Chu sequence, a network-level pilot base sequence a=[a ₀ , a ₁ ,...,a _τ-1 ] is generated, where τ is the sequence length. where a _n is generated by the following formula:

n=0, 1, ..., τ-1 (1)

Where i is the imaginary number unit, N is the control parameter, which should be an integer that is relatively prime to τ; a ₀ , a ₁ ,..., a _τ-1 are the elements of the sequence a, and the entire sequence is a pilot sequence

Step 2: Generate cell-level base sequences

According to the multiplexing factor in the topology structure of the cell in the network deployment, the number of cell-level base sequences to be generated is set as L. Considering that the cell-level base sequence can be reused in cells outside the first interference circle, generally L is taken as 3 or 7.

For the jth cell, the phase offset is performed on the basis of the network-level base sequence to generate the cell-level base sequence ψ _j of the cell, and the nth item of the sequence is calculated according to the following formula:

n=1, 2, . . . τ (2)

Where q _j is the phase offset control parameter of the cell, which cannot be an integer multiple of τ, and cannot be equal to j.

Step 3: Generate user-level pilot sequences

On the basis of the cell-level pilot base sequence, the user-level pilot sequence of the cell is obtained through cyclic shift. That is, assuming that the number of users that can be accommodated in a single cell is K, the kth user pilot sequence (k=1,2,...,K) of the jth cell is:

ψ _jk ＝〈 ψ _j 〉 _k-1 (3)

Among them, 〈·〉 _k means that the vector is cyclically shifted to the left by k bits.

Step 4: Calculate the statistical cross-correlation value between each user pilot sequence and each pilot sequence of the neighboring cell

Firstly, determine the maximum offset S when calculating the cross-correlation of pilot sequences between adjacent cells. The size of S can be configured according to the cell coverage scenario, that is, S is equal to the average propagation delay from the base station to the edge of the cell in the coverage scenario divided by the time domain length of a single symbol in the pilot sequence, or it can be set to 0 for simplicity (i.e. only 0-valued cross-correlations are considered).

For the constructed L cells × K users = LK user-level pilot sequences, calculate the statistical average of the cross-correlation values between the single user pilot and all user pilots outside the cell, that is, for the first For the kth user pilot of cell j, its statistical cross-correlation value is:

Where ccor _s (a, b) represents the cross-correlation value between sequence a and b with offset s.

Step 5: Sort and group the user pilot sequences in each cell according to their statistical cross-correlation values

It is set that the K user sequences in each cell can be divided into 2 groups, which are the center group and the boundary group. The ratio of the number of sequences in each group to the total number of user sequences in the entire cell can be configured according to needs, and the default is 1/2 , namely K/2 pieces.

Then, sort the K user sequences in each cell according to the statistical cross-correlation values calculated in step 4 from large to small, the front K/2 user pilots belong to the central group, and the following K/2 user pilots belong to for the boundary group.

Steps 1 to 5 are pre-calculated. Steps 6-9 are the process of user pilot allocation in the actual network.

Step 6: Determine the pilot group to which the user belongs

The user's pilot group is determined according to the current large-scale fading characteristics of each actual user in the network.

Central group: when the average value of large-scale fading from a user to all neighboring base stations that have a relationship with it is higher than the set threshold, or the dispersion of large-scale fading to each neighboring base station is smaller than a certain threshold, then the The user currently belongs to the central group.

Boundary group: when the dispersion of large-scale fading from a user to all neighboring base stations with which it has a neighboring cell relationship is greater than a certain threshold, or when the large-scale fading to a certain neighboring base station is greater than a certain threshold, the user currently belongs to for the boundary group.

Step 7: User Pilot Allocation

According to the group to which the user belongs, an unassigned pilot is randomly selected from the corresponding user sequence group of the cell to which the user belongs and assigned to the user.

Step 8: Channel estimate update

The allocation keeps M frames unchanged (synchronized with large-scale fading updates), performs channel estimation (single-cell MMSE or multi-cell MMSE) every frame, and updates the large-scale fading from each user to each neighboring base station after every M frames.

Step 9: User Pilot Update

According to the updated large-scale fading, the pilot groups to which users in the whole network belong are re-determined, and the pilots are randomly assigned in the group. Even if there is no change in the cell to which the user belongs and the pilot group to which it belongs, an unassigned pilot must be randomly selected from the corresponding user sequence group in the cell to which the user belongs (based on the idea of interference randomization).

The pilot sequence generated according to the above method has the following characteristics, that is, the plurality of cell-level base sequences generated by a network-level base sequence are orthogonal; and all user sequences generated by a cell-level base sequence are orthogonal It is also orthogonal, but the user sequences of different cells cannot be guaranteed to be orthogonal.

If the large-scale fading (path loss and shadow) is known, the signaling overhead when updating the pilot allocation is not considered.

Claims (2)

1. A pilot frequency allocation method, characterized in that: the method comprises the steps, Step 1: Generate network-level base sequences According to the generation method of the Chu sequence, a network-level pilot base sequence a=[a ₀ , a ₁ ,...,a _τ-1 ] is generated, and τ is the sequence length; where a _n is generated by the following formula: Where i is the imaginary number unit, N is the control parameter, which should be an integer that is relatively prime to τ; a ₀ , a ₁ ,..., a _τ-1 are the elements of the sequence a, and the entire sequence is a pilot sequence ; Step 2: Generate cell-level base sequences According to the reuse factor in the topological structure of the cell in network deployment, the number of cell-level base sequences to be generated is set to be L; considering that the cell-level base sequences can be reused in cells outside the first interference circle, generally L is taken as 3 or 7; For the jth cell, the phase offset is performed on the basis of the network-level base sequence to generate the cell-level base sequence ψ _j of the cell, and the nth item of the sequence is calculated according to the following formula: Where q _j is the phase offset control parameter of the cell, it cannot be an integer multiple of τ, and cannot be equal to j; Step 3: Generate user-level pilot sequences On the basis of the cell-level pilot base sequence, the user-level pilot sequence of the cell is obtained by cyclic shift; that is, assuming that the number of users that can be accommodated in a single cell is K, the k-th user pilot sequence of the j-th cell is : ψ _jk ＝< ψ _j > _k-1 (3) Wherein <·> _k means that the vector is cyclically shifted to the left by k bits, k=1, 2,..., K; Step 4: Calculate the statistical cross-correlation value between each user pilot sequence and each pilot sequence of the neighboring cell First determine the maximum offset S when calculating the cross-correlation of pilot sequences between adjacent cells; the size of S can be configured according to the cell coverage scenario, that is, S is equal to the average propagation delay from the base station to the edge of the cell in the coverage scenario divided by the pilot The time-domain length of a single symbol in the frequency sequence can also be set to 0 for simplicity; For the constructed L cells × K users = LK user-level pilot sequences, calculate the statistical average of the cross-correlation values between the single user pilot and all user pilots outside the cell, that is, for the first For the kth user pilot of cell j, its statistical cross-correlation value is: Where ccor _s (a, b) represents the cross-correlation value between sequence a and b with an offset of s; Step 5: Sort and group the user pilot sequences in each cell according to their statistical cross-correlation values It is set that the K user sequences in each cell can be divided into 2 groups, which are the center group and the boundary group. The ratio of the number of sequences in each group to the total number of user sequences in the entire cell can be configured according to needs, and the default is 1/2 , namely K/2 pieces; Then, sort the K user sequences in each cell according to the statistical cross-correlation values calculated in step 4 from large to small, the front K/2 user pilots belong to the central group, and the following K/2 user pilots belong to is the boundary group; Steps 1 to 5 are pre-calculated; steps 6 to 9 are user pilot allocation processes in the actual network; Step 6: Determine the pilot group to which the user belongs Determine the user's pilot grouping according to the current large-scale fading characteristics of each actual user in the network; Central group: when the average value of large-scale fading from a user to all neighboring base stations that have a relationship with it is higher than the set threshold, or the dispersion of large-scale fading to each neighboring base station is smaller than a certain threshold, then the The user currently belongs to the central group; Boundary group: when the dispersion of large-scale fading from a user to all neighboring cell base stations with which it has a neighbor cell relationship is greater than a certain threshold, or when the value of large-scale fading to a certain neighboring cell base station is smaller than a certain threshold, the user Currently belongs to the boundary group; Step 7: User Pilot Allocation According to the group to which the user belongs, randomly select an unassigned pilot from the corresponding user sequence group of the cell to which the user belongs to assign to the user; Step 8: Channel estimate update The allocation remains unchanged for M frames, channel estimation is performed for each frame, and the large-scale fading from each user to each adjacent base station is updated after each M frame; Step 9: User Pilot Update According to the updated large-scale fading, re-determine the pilot group to which the users of the entire network belong, and randomly assign pilots in the group; Randomly select an unassigned pilot from the corresponding user sequence group of the cell to which it belongs and assign it to the user; The pilot sequence generated according to the above method has the following characteristics, that is, the plurality of cell-level base sequences generated by a network-level base sequence are orthogonal; and all user sequences generated by a cell-level base sequence are orthogonal It is also orthogonal, but the user sequences of different cells cannot be guaranteed to be orthogonal. 2. A pilot allocation method according to claim 1, characterized in that: if the large-scale fading, that is, the path loss and the shadow, are known, the signaling overhead when updating the pilot allocation is not considered.