The present invention relates to communication networks and, in particular, to a multiple-input multiple-output based space division multiple access technology.
MIMO technology (Multiple-Input Multiple-Output, multiple-input multiple-output) allows multiple antennas simultaneously transmit and receive multiple spatial streams. By using the MIMO technology, the channel capacity can be increased, in the meantime, the reliability of a channel can be improved, and the error rate can be reduced. The SDMA (Space Division Multiple Access, space division multiple access) is a newly developed multiple access technology, which use spatial segmentation to constitute different channels. Because the SDMA technology has the potential to greatly improve the capacity and performance of wireless communication systems, it has caused widespread concern from researchers.
However, when the SDMA technology is applied to the MU MIMO scenario, the coordination between users becomes a key issue. Taking a cellular network as an example, in a downlink transmission, base stations simultaneously transmit signals to users. Therefore, part of signals received by each user is inter-user interference caused by signals of other users. Although, in theory, the user can utilize multi-user detection technology to eliminate multi-user interference (Multi-User Interference, MUI), given the demand of low power consumption, low complexity and low cost of the user, the MUI is expected to be eliminated at the base station side. Under the premise that channel state information (Channel State Information, CSI) is obtained by the base station side, the inter-user interference can be eliminated by adopting the linear or non-linear pre-coding technology. However, in a MU (Multi-User, multi-user) MIMO system, in order to obtain SDMA gain, spacial multiplexing user number on a given frequency and within a given time slot is limited. In addition, if the users which are served at the same time have great spatial correlation, the transmission performance would be reduced greatly. Therefore, it is need to utilize spatial scheduling algorithms to process the user selection in SDMA, so as to maximize the transmission performance of the system.
In the present ProSched spatial scheduling algorithms, the pre-coding process is interpreted as a mapping, thereby avoiding calculating pre-coding matrixes of all of the possible combinations of users. However, the original intention of the ProSched algorithms is to be applied to a single carrier system. Although an extended method which can be applied to the multi-carrier system can also be realized, this extended method is not applicable to IEEE 802.11ac system.
There is another existing method for performing the SDMA wireless communication with at least two users simultaneously. According to that patent application, all users are pre-sequenced in accordance with a certain rule, then all possible combinations of users are generated from these pre-sequenced users, where each combination includes at least two users. In this mechanism, it is assumed that each user has only one antenna. After the base station selects all combinations of users from K user sets, where each combination of users includes U users, the base station calculates the capacity of the jth combination of users as follows. First, the base station obtains a weight vector of the jth combination of users:
W
ZF
=H
H(HHH)−1=[w1w2 . . . wU]
where, WZF represents a zero-forcing weight matrix, H represents a joint channel matrix of the jth combination of users. Then, the base station obtains SNR of the uth user in the jth combination of users group with the following formula:
where, xu represents a symbol transmitted by the uth user, wu is a zero-forcing weigh vector of the uth user, σu2 is noise variance. Therefore, based on the SNR of the user, total capacity of the jth combination of users Cjtotal may be calculated according to the following formula:
The method uses the calculated capacity to perform scheduling.
However, the mechanism in the prior art is provide for the case that there is only one antenna at the UT side. Furthermore, the pre-coding algorithm is limited to the zero-forcing method, and how to use the scheduling mechanism and the present pre-coding algorithms jointly is not provided. When calculating the capacity according to the method, in which the zero-forcing (ZF) capacity or SNR of each combination of users is calculated and then a combination of users with the maximum ZF capacity or SNR is selected, calculating the weight matrix by adopting the zero-forcing method is introduced, however, the architecture of the MU MIMO transmission is not considered. In addition, it should be noted that the index in this method is designed only for the single carrier system or for one sub-carrier of the multi-carrier system, and could not be extended directly to be applied to the multi-carrier system.
Embodiments of the present invention provide a new method for an SDMA transmission method to process selection of SDMA users in spatial scheduling, so as to maximize the transmission performance of the system.
At the first aspect, embodiments of the present invention provide a method for scheduling a SDMA user group in an MU MIMO system. The method includes the following steps:
calculating a projection matrix of a channel information matrix of a user on a joint channel matrix, where the joint channel matrix is a joint channel information matrix of other users in a user group of which the user is a member;
calculating a capacity metric of the user group according to the projection matrixes;
determining, according to capacity metrics of a plurality of user groups, a user group for which a SDMA transmission needs to be provided.
At the second aspect, embodiments of the present invention provide a base station which is configured to perform scheduling of user group in an SDMA system. The base station includes:
a projection matrix calculating module, configured to calculate a projection matrix of a channel information matrix of a user on a joint channel matrix, where the joint channel matrix is a joint channel information matrix of other users in a user group of which the user is a member;
a capacity metric calculating module, configured to calculate a capacity metric of the user group according to the projection matrix;
a user group selecting module, configured to determine, according to capacity metrics of a plurality of user groups, a user group for which a SDMA transmission needs to be provided.
In embodiments of the present invention, the spatial correlation of the users is reflected by the projection matrixes of users on different sub-carriers, the user group is selected by calculating the projection matrixes, thereby the MUI of the terminal user is effectively eliminated at the base station side. Embodiments of the present invention can also reduce the energy loss, which is caused by the spatial correlation between the users, in the uplink direction.
Referring to the drawings, the following describes embodiments of the present invention more concretely, where:
The base station has the information of channels between the MT transmitting antenna and the receiving antennas of a single user terminal. It should be noted that, in this description, a user and a user terminal have the same meaning.
At the base station side, a transmitting signal to be transmitted to the ith user is defined as a ri-dimensional vector, xi, herein i=1, 2, . . . , K, ri is number of data stream transmitted to the ith user. A vector of K users may be expressed as x=[x1T x2T . . . xKT]Tεr×1, herein
Each transmitting signal needs to be pre-coded before being transmitted. A pre-coding matrix for K users may be expressed, by adopting the form of a joint pre-coding matrix, as: F=[F1 F2 . . . FK]εM
M
The transmitting signal for the ith user may be transmitted through MT transmitting antennas by adopting the SDMA mode, and then be received by one of or a plurality of the MR
The joint channel matrix of K users is expressed as follows: H=[H1T H2T . . . HKT]εM
At the user terminal side, the received signal is decoded. A decoding matrix for the user terminal of the ith user is
The decoding matrix for K users may be expressed, by adopting the form of a joint block-diagonal decoding matrix (joint block-diagonal decoding matrix), as:
Therefore, the receiving vector of K users may be expressed, by adopting the form of joint receiving vector, as:
y=D·(H·F·x+n) (1)
Herein, y=[y1T y2T . . . yKT]εr×1, yiε
r
M
In the MU MIMO downlink data transmission, MUI (multi-user interference) on the user terminals is eliminated by using the pre-coding algorithm, and the computational complexity is reduced. A solution to the above problems is to determine, by using the spatial scheduling, which users and how many users can be determined to form an SDMA communication group for being served at the same time and at the same frequency.
According to embodiments of the present invention, the pre-coding process procedure is interpreted as a new orthogonal projection, and the pre-coding process is approximated by the orthogonal projection. In this application, this method is called EMC-ProSched (Efficient Multi-carrier, efficient multi-carrier) user grouping and scheduling method. This method is described in detail in the following.
Firstly, the channel matrix of the ith sub-carrier of the ith user is defined as Hi,j. In the user group S which includes N users, the rate of the jth sub-carrier of the ith user is estimated based on a method for calculating the minimum capacity limit of each sub-carrier of the ith user:
C
i,j=log2(1+σn−2∥Hi,jFi,j∥F2), (2)
Herein, σn2 represents the noise power on each sub-carrier bandwidth of the receiver, Fi,jεM
Then, the diagonal matrix which includes the square root of each spatial stream allocated to the ith user is extracted. The remaining part of the pre-coding matrix is denoted by Mi,jεM
F
i,j
=M
i,j
G
i,j.
The objective of the spatial scheduling algorithm with low complexity is to avoid calculating the pre-coding matrixes of all user groups. Before pre-coding, the eigenvalues distribution of an equivalent channel Hi,jFi,j is not known. Therefore, it is difficult to obtain the optimal power allocation scheme when selecting users. Assuming that the total transmitting power is divided equally by users and, meanwhile, the power adopted by each spatial streams is equal, then the corresponding power diagonal matrix Gi,j may be expressed as:
G
i,j=√{square root over (PT/(N·ri)I)} (3)
Herein PT is the transmission power allocated to each sub-carrier. N is number of users in the user group, ri is number of spatial streams of the ith users. Compared to using optimal power allocation strategy, when using this equal power allocation strategy, the system performance is reduced.
Substitute equation (3) into equation (2), and the estimated transmission rate of the jth sub-carrier of the ith user in the user group S is obtained:
The Log 2 function is a monotonically increasing function. Since only estimating levels of spatial correlation of different user groups according to the metrics is needed in embodiments of the present invention, therefore, equation (4) can be simplified to the following metric:
Hi,jMi,j is called an equivalent channel matrix.
Therefore, the scheduling metric of the ith user in the user group S may be defined as the weighted sum of the estimated rate (capacity) of the user on all sub-carriers:
Herein, NSD is number of data sub-carriers, δj is a weight factor, j=1, 2, . . . , NSD. In an example, all δj are 1, which means that weighting is not performed. In another embodiment, the weight factor δj may be configured as a step function (step function):
Herein ε is a threshold. Under the effect of the weight factors, a sub-carrier, of which the frobenius of channel is small, will no longer be considered in the calculation of the metric of the user. This can guarantee that the user, of which only some sub-carriers have good channel quality while other sub-carriers suffer deep fading, will not be selected.
In the case that BD is used as the pre-coding algorithm, the frobenius of the equivalent channel matrix is equal to the frobenius of the projection channel, that is:
∥Hi,jMi,j∥F2=∥Hi,j{tilde over (P)}i,j∥F2
Herein, the projection channel of the jth sub-carrier of the ith user is {dot over (H)}i,j=Hi,j{tilde over (P)}i,j, in which {tilde over (P)}i,j is the projection matrix of the matrix H. For other linear pre-coding algorithms, i the two sides of the above equation (7) are approximately equal. Therefore, the impact of the spatial correlation between users is taken into account by the frobenius (Frobenius) of the projection channel ∥Hi,j{tilde over (P)}i,j∥ also.
Therefore, the matrix Mi,j in equations (5) and (6) may be replaced by {tilde over (P)}i,j. The scheduling metric of the ith user in the user group S can be expressed as:
It can be seen from the above equation, the scheduling metric of the user can be calculated according to the projection matrix of the channel matrix Hi, and thus the selection of users is allowed, and the purpose of eliminating multi-user interference can be achieved.
Accordingly, in embodiments of the present invention, based on the relationship between the projection matrix of the users on different sub-carriers and the spatial correlation between the users, the projection matrix of each user on different sub-carriers is calculated, the frobenius of the product of the channel matrix and projection matrix of the user is solved and used as the scheduling metric to perform the selection of a user group, and then the SDMA transmission is performed for the selected user group. For a user group with high spatial correlation, the frobenius (Frobenius) of the projection channel matrix of each user in the user group is very small; the pre-coding efficiency of the user group will be greatly reduced, thereby the data rate will be also significantly reduced. According to the present invention, the sum of the metrics of this user group will be very small, which means that the possibility that this user group is selected is relatively low.
Although, in the foregoing, the BD algorithm is taken as an example to described the specific embodiments of the present invention, the scheduling criterion proposed by the EMC-ProSched may also be combined with other linear pre-coding algorithm (such as LP-GMD, LoCCoBF) besides the BD algorithm, to select the most suitable user group.
In an embodiment, according to IEEE 802.11 ac specification, an SDMA group is valid for at least one transmission packet. It can be seen from the simulation of the channel model D in 802.11ac, in the view of time dimension, the channel conditions do not change significantly when one packet is transmitted. Therefore, when performing the spatial scheduling algorithm, only the sub-carriers of one OFDM symbol needs to be considered.
In an embodiment, in order to avoid solving the pre-coding matrixes of all possible user groups in advance, a projection matrix of the jth sub-carrier of the ith user in the user group may be calculated by using the mode of the orthonormal basis of the signal space on the jth sub-carrier of the ith user. The approximate projection matrix {tilde over (P)}i,j is acquired by using the following equation:
{tilde over (P)}
i,j=(Pi,j, . . . ,Pi−1,j,Pi+1,j, . . . ,Ps,j)p,p→∞
Herein, p is a projected order. {tilde over (P)}i,j may also be referred to as the joint projection matrix of the jth sub-carrier of other users except the projection matrix of the jth sub-carrier of the ith user. The projection order P may be selected as 1˜3, which is sufficient. {tilde over (P)}i,j may be calculated by using the orthonormal basis of the signal space on the jth sub-carrier of the ith user,
P
i,j
=I−B
i,j
B
i,j
H. (9)
In another embodiment, a singular value decomposition (singular value decomposition (SVD)) method is adopted to obtain the projection matrix {tilde over (P)}i,j. That is, the singular value decomposition is performed to the joint channel matrix of the other users on the jth sub-carrier except the ith user, thereby the projection matrix of the jth sub-carrier of the ith user in the user group is calculated.
Specifically, the channel matrix {tilde over (H)}i,j, of which the rank is {tilde over (L)}i,j, is the joint channel matrix of the other users on the jth sub-carrier except the ith user, which is defined as follows:
{tilde over (H)}
i,j
=[{tilde over (H)}
i,j
T
. . . {tilde over (H)}
i−1,j
T
{tilde over (H)}
i+1,j
T
. . . {tilde over (H)}
K,j
T]T.
By performing SVD to the channel matrix {tilde over (H)}i,j, the following equation can be obtained:
{tilde over (H)}
i,j
=Ũ
i,j{tilde over (Σ)}i,j[{tilde over (V)}i,j(1){tilde over (V)}i,j(0)]H (10)
Herein {tilde over (V)}i,j(1) contains the first {tilde over (L)}i,j right singular vectors (right singular vectors); {tilde over (V)}i,j(0) contains following (MT−{tilde over (L)}i,j) right singular vectors, these right singular vectors constitute the orthonormal basis of null space (null space) of {tilde over (H)}i,j, {dot over (H)}i,j represents the equivalent channel matrix of the jth sub-carrier of the ith user after the MUI is eliminated, which is obtained by multiplying a channel matrix of the jth sub-carrier of the ith user with a projection matrix {tilde over (P)}i,j, that is
{dot over (H)}
i,j
=H
i,j
{tilde over (P)}
i,j
Herein, {tilde over (P)}i,j may be expressed as
{tilde over (P)}
i,j
={tilde over (V)}
i,j
(0)
{tilde over (V)}
i,j
(0)H (11)
In step 312, for each of all possible user groups, calculate a sum of the metrics of users in a user group, and select the user group S1opt, corresponding to the maximum sum of metrics.
In step 320, add a user, which is selected from the remaining users, into the user group SK
In step 322, re-calculate the metric of each user in the new user group, and calculate the sum of the metrics for the new user group accordingly.
In step 324, select a best user group SK
In step 330, compare number of users in the user group with the total number of users in the system K. If Kg<K, return to step 320. If Kg=K, compare the sums of the metrics of user groups with different user number (such as S1opt, S2opt, . . . , SKopt). According to the metrics of the plurality of user groups, select the user group. In an embodiment, select a user group corresponding to the maximum sum of metrics, then number of users in the selected SDMA user group, is the optimal size of an SDMA user group.
In the preceding step 312, a variety of methods may be used to determine possible user groups. In an embodiment, a tree-based sorting algorithm may be used to add a user or delete a user from a user group, so as to determine the possible user groups. The tree-based sorting algorithm is described in detail as following. Assuming there are, for example, 5 users, i.e., users 1-5, in the MU MIMO system. First, under the premise that a group contains only one user, calculate the sum of metrics for each group, where the algorithm for the calculation may be the SVD algorithm; as a result, the user group, of which user 1 is a member, is determined to be the optimal user group. Next, add user 2, of which the metric is greater, to the user group of which user 1 is a member, then re-calculate the sum of metrics for each group (of course, the user groups of which users 3-5 are a member respectively have not changed, so they are not necessary to be re-calculated), and then determine an optimal user group (user 1, user 2) is determined, where the algorithm for the calculation may be the SVD algorithm or the algorithm which combines equation (9) and equation (8). Next, according to values of the metrics, add user 4, user 3 and user 5 into the user group successively. Finally, compare the sums of metrics for the user groups in each round, and determine the optimal user group.
The scheduling according to the above steps, at the initial stage, only needs to perform SVD decomposition KNSD times, and the main complexity of performing pre-coding by using SVD decomposition method concentrates on the calculation amount of the SVD decomposition. The complexity of performing SVD to a M*N matrix is min(MN2,M2N). Therefore, the EMC-ProSched algorithm may be called low complexity algorithm, because the EMC-ProSched algorithm reduces, comparing with the conventional user selection algorithms, times of performing the SVD decomposition, especially for the multi-carrier systems. For example, for a system having 16 users, number of data sub-carriers NSD is 234, the size of the user group is set to be 4, the times of SVD decomposition needed by the EMC-ProSched algorithm is 234×16=3744. However, for traditional scheduling algorithms, the same scheduling task need 234×((16−1)×2+(16−2)×3+(16−3)×4)=29016 times SVD decomposition.
Therefore, compared with the ProSched scheduling method, the EMC-ProSched has lower computational complexity, and needs less signaling overhead. In addition, the scheduling method provided by the present patent application may be combined more flexibly with the pre-coding algorithms, and there is no limit to number of antennas of the user terminal.
The present invention is applicable not only to the MU MIMO downlink system in which SDMA is combined with OFDM, but also to the MU MIMO downlink system in which SDMA is combined with OFDMA. It should be noted that, the multiple access of OFDMA is achieved by allocating different sub-carriers to different users. For example, all users transmit simultaneously, and each user adopts different sub-carrier sets. When SDMA and OFDMA are jointly used, each sub-carrier set is used by a group of users, rather than by one. Therefore, when a user group in which the users are served simultaneously on the same sub-carrier set, the scheduling algorithm EMC-ProSched may be further performed to the selected user group.
Furthermore, the EMC-ProSched algorithm may be extended to be applicable to the uplink multi-user MIMO transmission.
As a method for rate estimation, the EMC-ProSched algorithm may be used jointly with the fairness algorithm directly, so as to achieve an equitable space assignment in terms of time and throughput and the like. In an example, when proportional criteria is adopted, the rate estimation for the ith user in the user group S may be obtained by normalizing its average metric in a long time:
γi(S)=ηi(S)/
Herein
Because of the normalization of the average rate in a relatively long time, the proportional fair scheduling can solve the near-far problem to a certain extent.
A capacity metric calculating module 504 calculates a capacity metric of the user group according to the projection matrix. In an example, the capacity metric calculating module performs fairness algorithm calculation to the capacity metric of the user group. In an embodiment, the capacity metric calculating module calculates a capacity metric of each sub-carrier of each user in the user group according to the projection matrix of the user; weights the capacity metric of each sub-carrier of each user in the user group; calculates a capacity metric of each user in the user group and the capacity metric of the user group, based on the weighted capacity metric of each sub-carrier of each user.
a user group selecting module 506 selects a user group according to the capacity metrics of a plurality of user groups.
Then the base station transmits, in accordance with the provision of SDMA, the data of each user in the selected user group through several antennas, respectively, after performing the series of process of coding, modulation, filtering, and so on.
In an example, the base station further includes a module which adds a user to or deletes a user from the user group by adopting a tree-based algorithm to obtain the plurality of user groups.
In an example, the MU MIMO system is an OFDMA system. The base station selects the users which are served at the same time and on the same sub-carrier set. For this kind of user group, the base station performs user group scheduling, that is, makes the projection matrix calculating module, the capacity metric calculating module, and the user group selecting module performs the corresponding operation to the user group in which the users are served at the same time and on the same sub-carrier-set.
In an example, the base station includes a transmitting module which provides SDMA transmission in downlink direction to the users simultaneously in the selected user group.
In an example, the base station includes a transmitting module which transmits the pre-coding matrix determined according to the selected user group to the terminal. The terminal performs data encoding to each user in the user group after receiving the pre-coding matrix, and transmits the encoded data in uplink direction.
Referring to
S610. Calculate a projection matrix of a channel information matrix of a user on a joint channel matrix, where the joint channel matrix is a joint channel information matrix of other users in a user group of which the user is a member.
For example, a projection matrix of the ith sub-carrier of the ith user in the user group is calculated by using the orthonormal basis of signal space on the ith sub-carrier of the ith user, herein i, j are positive integers respectively. Or, singular value decomposition is performed to the joint channel matrix of the other users except the ith user on the jth sub-carrier, and then a projection matrix of the ith sub-carrier of the ith user in the user group is calculated, herein i, j are positive integers respectively.
S620. Calculate a capacity metric of the user group according to the projection matrix.
In the calculation, a fairness algorithm calculation is performed to the capacity metric of the user group. The specific method may include: calculating a capacity metric of each sub-carrier of each user in the user group according to the projection matrix of the user; weighting the capacity metric of each sub-carrier of each user in the user group; calculating a capacity metric of each user in the user group and a capacity metric of the user group, based on the weighted capacity metric of each sub-carrier of each user. Or, the method may include calculating the frobenius of the product of the channel matrix of the user and the projection matrix of the user, and taking the frobenius as the capacity metric of the user.
S630. Determine, according to capacity metrics of a plurality of user groups, a user group for which a SDMA transmission needs to be provided.
In the implementation, a tree-based algorithm may be used to add or delete a user so as to obtain the plurality of user groups, and the capacity metrics of the plurality of user groups is further calculated.
In this embodiment, if the MU MIMO system is an OFDMA system, then before the step of calculating the projection matrix of a channel information matrix of each user in the user group on the joint channel matrix of the other users in the same user group, the method includes: selecting users which are served at a same time and on a same sub-carrier set, and constituting the user group.
In order to verify the performance of the scheduling algorithms provided in the present patent, a simulation is made by using MATLAB. The configuration and parameter setting of the simulation are in accordance with the specifications of the scenario 2 in IEEE 802.11ac [IEEE451].
It should be noted that the foregoing embodiments of the present invention have been described with a combination of the MU MIMO system, but the present invention is not limited thereto. For example, the present invention may also be applicable to a MIMO system which includes a single-input single-output (SISO) user. Furthermore, in the foregoing embodiments, the present invention calculates the capacity metric of the user group according to the projection matrix on all the sub-carriers of each user. However, it is also feasible to group sub-carriers (for example, the sub-carriers have a high correlation between each other are grouped into a same group), select a sub-carrier from the sub-carrier groups and calculate the projection matrix on the selected sub-carrier of each user.
Obviously, the invention described herein may have many variations, and such variations are not regarded as a departure from the spirit and scope of the invention. Therefore, all modifications apparent to the persons of ordinary skill in the art are included within the scope of claims of the present invention.
Number | Date | Country | Kind |
---|---|---|---|
201110262687.2 | Sep 2011 | CN | national |
This application is a continuation of International Application No. PCT/CN2012/081079, filed on Sep. 6, 2012, which claims priority to Chinese Patent Application No. 201110262687.2, filed on Sep. 6, 2011, both of which are hereby incorporated by reference in their entireties.
Number | Date | Country | |
---|---|---|---|
Parent | PCT/CN2012/081079 | Sep 2012 | US |
Child | 14199471 | US |