Summary of the invention
Technical problem: in order to overcome the deficiency of existing pilot tone dispatching algorithm, the invention provides a kind of based on user's spacing
From dispatching algorithm, can reach to approach, with relatively low complexity, the scheduling performance that traversal is optimum.
Technical scheme: found by analysis, uses least-mean-square error algorithm to carry out the situation of channel estimation at receiving terminal
Under, when distance tends to infinity to the user distributing same pilot tone each other, mean square estimation difference will be close to zero.Based on this,
This dispatching algorithm reduces the most greatly mean square estimation difference by making the spacing being assigned to often organize the user under pilot tone.Adjust
The basic thought of degree: choose successively and be assigned to often organize the user under pilot tone;When each iteration, choose distance each other and to the greatest extent may be used
While user that can be big so that remaining user disperses as far as possible.
Pilot distribution method based on user's spacing under a kind of large-scale distributed antenna system, it is characterised in that choosing
Go out the user being in a disadvantageous position as key user after, get rid of P-1 the user closest with it, by remaining users recurrence use
This distribution method is divided into P group, selects best one group and the user being in a disadvantageous position is assigned under same pilot group, under then carrying out
The distribution of one group of pilot tone, wherein, P is pilot number.
The concrete steps of pilot tone distribution:
Step 1: choose key user ukey:
Step 2: distribution comprises key user ukeyA pilot group, concrete grammar is:
With u in eliminating UkeyP-1 closest user, by remaining P (Kpilot-1) individual user calls Recursive schedule calculation
Method is divided into P group, often group (Kpilot-1) individual user (is i.e. P (K by total number of userspilot-1), pilot number is that the situation of P is carried out point
Joining, concrete recurrence performs step and sees below), it is designated asWherein, KpilotFor often organizing the use of pilot tone distribution
Amount;
Select optimum one groupWith ukeyIt is assigned to together under same group of pilot tone, constitutes
Step 3: repeat step 1 and step 2, be sequentially completed the distribution of a pilot group.
The choosing method of key user is:
Step 1-1: distance threshold d is setth, obtain user and gather the user bunch all corresponding to this threshold value in U, take wherein
Bunch of C that number of users is most, if its number of users is KC;
Step 1-2: if number of users K in bunchCNot less than pilot number P, the central user choosing bunch C is key user ukey:
In formula, D (C) is the distance matrix in user bunch C between user, [D (C)]ijRepresent distance matrix the i-th row jth row first
Element, i, j corresponding user respectively gathers in U i-th, j user;
Otherwise, choose whole user gathering the central user of U is key user ukey:
Optimum one groupFor:
M () represents mean square estimation difference when one group of user is assigned to same pilot group.
The execution step of described Recursive schedule algorithm is:
Often distribute two user (K under group pilot tonepilot=2), time, comprise the concrete steps that:
A, select key user u according to step 1key, get rid of P-1 the user closest with it, P user of residue
In select best one with key user ukeyIt is assigned under same group of pilot tone;
B, get rid of allocated 2 users, remaining users is selected key user again, repeat step a until will be all
User is distributed into P group;
Often distribute three user (K under group pilot tonepilot=3), time, comprise the concrete steps that:
Key user u is selected according to step 1key, get rid of P-1 the user closest with it, residue 2P user pressed
According to KpilotAssigning process when=2 is divided into P group, selects best one group and key user ukeyIt is assigned under same group of pilot tone;Row
Except allocated user, repeat this process until all users are divided into P group.
……
Often distribute S user (K under group pilot tonepilot=S) time, comprise the concrete steps that:
Key user u is selected according to step 1key, get rid of P-1 the user closest with it, will residue (S-1) P use
Family is according to KpilotAssigning process during=S-1 is divided into P group;One group that selects optimum is assigned to same group of pilot tone with key user
Under;Get rid of allocated user, repeat this process until all users are divided into P group.
It is optimum (in performance that the algorithm that pilot distribution method of the present invention uses can realize approaching traversal with relatively low complexity
Limit) estimation performance.
Detailed description of the invention
Below in conjunction with the accompanying drawings and instantiation, it is further elucidated with the present invention, it should be understood that this example is merely to illustrate the present invention
Rather than restriction the scope of the present invention, after having read the present invention, the those skilled in the art's various equivalences to the present invention
The amendment of form all falls within the application claims limited range.
First following description is done: in distributed extensive mimo system, total K single-antenna subscriber, constitute set U,
The each RAU of base station end is also equipped with single antenna, and RAU number is M.EPRepresent pilot power,Represent
Pth pilot frequency sequence, and meethk∈CM×1Represent the kth user channel vector to base station end, Y
∈CM×τIt is to receive signal matrix, N ∈ CM×τIt it is noise matrix.
The system model in uplink channel estimation stage:
At receiving terminal, reception signal is done relevant to known pilot frequency sequence, and defines
Mutually orthogonal in view of different pilot frequency sequences, then have
WhereinThe estimated value obtained by MMSE estimation and the mean square estimation difference of correspondence are
It is i-th element r on diagonal matrix, and diagonalk,iRepresent the user k that i-th RAU receives
Mean power.Our research has a problem in that the user that how to distribute in U is to different pilot group { Up, so that overall
Mean square estimation difference is minimum, and mathematical notation is
Distance matrix D (U)=[d of U is collected based on userij]i,j∈UDefine two class users:
It is in the user of center:
Owing to there is no the positional information of user, it is contemplated that compare boundary user and be in user and other users of center
Ultimate range can be less, therefore had the definition of above central user based on distance.
User bunch.Set a distance threshold dth, it is such a collection for one group of user U, user bunch C thereon
In conjunction bunch, the distance between any two user u and u ' is not more than threshold value dth, bunch outer any user with bunch at least one
Distance between user is more than threshold value dth.Mathematical notation is
Our algorithm chooses the user's group being assigned to often organize under pilot tone successively.For the user bunch that number of users is more, should
In avoiding as far as possible bunch, user is assigned to same pilot group, therefore in the presence of such bunch, and the center that we preferentially select bunch
User, and select the one group user distant with it to be assigned to same pilot group.And on whole user's collection to be allocated
Central user, owing to the number of users distant with it comparatively speaking can be fewer, when there is not the user bunch of relatively multi-user
Also should pay the utmost attention to.
First introducing the situation that number of users K=| U | is pilot number P integral multiple, the most often group pilot tone distributes same number Kpilot=
The user of K/P.For general situation, need only the most slightly improve, after have specific description.
Number of users is pilot tone allocation step during pilot number integral multiple, as shown in Figure 1:
Step 1: obtain the Distance matrix D (U) between user, and p=1 is set;
Step 2: cluster.Distance threshold d is setth, obtain user and gather the user bunch all corresponding to this threshold value in U,
Take bunch of C that wherein number of users is most, if its number of users is KC。
Step 3: choose key user ukey.Selection standard:
If number of users K in bunchCNot less than pilot number P, the central user choosing bunch C is key user ukey:
Otherwise, choose whole user gathering the central user of U is key user ukey:
Step 4: select one group of user the most distant and ukeyIt is assigned under same pilot group.Get rid of in U
With ukeyP-1 closest user, by remaining P (Kpilot-1) individual this dispatching algorithm of user's recursive call is divided into P group, often
Group (Kpilot-1) individual user, is designated as(note: if there is multiple number of users in step 2 more than pilot tone
Number bunch, then in addition to selected bunch C, these bunches remaining can progressively be spread out in this step)
Select optimum one group and ukeyIt is assigned to together under pth group pilot tone, constitutes
M () represents mean square estimation difference when one group of user is assigned to same pilot group.
Step 5: gather from user and get rid of the most the allocated user UUnallocated pilot number P=P-
1, remaining users number K=K-Kpilot, reacquire Distance matrix D (U), repeat step 2 to 5 and carry out next group pilot tone p=p+1
Distribution.
In a practical situation, number of users K may not be the integral multiple of number of pilots P.When this happens, may be used
To be scheduling as follows:
Step 1: takeThen there is P1
+P2=P,
Step 2: distribution P2Individual containThe pilot group of individual user:
A, acquisition user gather the user u being in border in Uborder:
With u in B, eliminating UborderClosest P2-1 user, by remainingIndividual user is by the tune of a upper trifle
Degree algorithm is divided into P group, often groupIndividual user, is designated as
C, select from P group user obtained in the previous step optimum one group and uborderConstitute together
Delete user gather in U withCorresponding user, repeats step A, B, C until distributing P2Group user
Step 3: remaining user is divided into P by the dispatching algorithm of a upper trifle1Group, often groupIndividual user.
Actual range between user may be not readily available, and the concept of our definition " interference distance " is in order to substitute.Examine
Consider two users and share the mean square estimation difference of one of them user under a pilot tone situation:
Accordingly, the mean square estimation difference when not having pilot pollution to only have noise is
The impact that analysis pilot pollution causes:
Likewise it is possible to obtain the impact that another user is caused by pilot pollution
DefinitionWhat we can approximate thinks, Δ I12The least, the pilot pollution between user 1,2 is asked
Topic will be the least.Definition interference distance
So interference distance is the biggest, and the pilot pollution problem between user will be the least, and between user, actual range is dirty to pilot tone
As the impact of metachromia energy.
At K=14, during P=4,2 groups of pilot tones are had to distribute 4 users, 2 groups of pilot tone 3 users of distribution.First distribution is containing 4
The pilot group of user:
Step 1: obtain user and gather the user u being in border in Uborder:
Step 2: with u in eliminating Uborder1 closest user, by remaining 12 users by the tune of a upper trifle
Degree algorithm is divided into 4 groups, and often 3 users of group, are designated asCalculate this four groups of users and u respectivelyborderGroup
The mean square estimation difference being combined, select corresponding mean square estimation difference minimum one group and uborderDistribute together at first group
Under pilot tone;
Step 3: get rid of allocated user, selects the user u being in border in 10 users of residueborder2, equally will
Remain 9 users and be divided into 3 groups, select best one group and boundary user uborder2It is assigned to together under second group of pilot tone;
Then it is divided into two groups by remaining 6 users, often 3 users of group.
Step 4: select the key user u in 6 userskey, get rid of a user nearest with it, and 4 use will be remained
K is pressed at familypilotAllocation algorithm when=2 is divided into 2 groups.Calculate this two groups of users and u respectivelykeyThe mean square estimation combined is by mistake
Difference, selects less one group and ukeyIt is assigned under the 3rd group of pilot tone;Remaining 3 users are assigned under the 4th group of pilot tone.
Emulate the distributing antenna system being equipped with single antenna based on each RAU end.Base station end 100 single antenna of configuration
RAU, each RAU are uniformly distributed on geographical position, neighbouring 4 RAU constitute one between square, and adjacent R AU away from
From for 100m, customer location stochastic generation.
It is simulation result during pilot number integral multiple (K=12, P=4) that Fig. 2 gives number of users.Fig. 2 (a), 2 (b) are respectively
Give channel estimation errors and average relation between user rate and signal to noise ratio.In figure, " Best " represents traversal optimum and calculates
Method, be take after the mean square estimation difference that all possible pilot allocation scheme travels through and compares different schemes the most equal
Estimation difference minimum a kind of in side's obtains, and this is the higher limit of performance;" Worst " represents and travels through worst algorithm, under performance
Limit value;" Greedy " represents greedy algorithm;" Loc " represents dispatching algorithm based on distance in this paper;" random " represents
Random schedule.It can be seen that whether channel estimation errors or average user speed, algorithm performance in this paper is the most permissible
Approach traversal optimal performance, and be better than greedy algorithm and Randomized scheduling algorithm.
It is not simulation result during pilot number integral multiple (K=14, P=4) that Fig. 3 gives number of users.In this case,
Greedy algorithm is also pilot group and the pilot group of 2 groups of 3 users that all users are divided into 2 groups of 4 users, and first distributes 4
The pilot group of 3 users is distributed after the pilot group of user.Owing to ergodic algorithm is the most complicated, corresponding emulation knot
Fruit is not given.It will be seen that location-based dispatching algorithm estimation difference performance when high s/n ratio is still substantially better than greedy
Greedy algorithm.
Fig. 4 gives the simulation result during obedience Poisson distribution of RAU position, it can be seen that algorithm performance in this paper depends on
It is so optimum.