CN103905097A - Distributed antenna system resource scheduling method with self-adaptive antenna selection - Google Patents

Abstract

The invention belongs to the technical field of communication, and particularly relates to a distributed antenna system resource scheduling method with self-adaptive antenna selection. A constrained model with L1 norms is established for the resource allocating problem in a distributed antenna system. The model is converted into a dual problem to be solved, the dual problem is divided into a plurality of independent sub-problems to be solved, and the optimal solution of the sub-problems is provided in the view of analytical geometry. The simulation result shows that the algorithm is high in convergence speed and superior to a central system in performance, most importantly, the number of selected ports is relatively small, but the corresponding performance loss is very small.

Description

The distributing antenna system resource regulating method of a kind of combining adaptive sky line options

Technical field

The invention belongs to communication technical field, be specifically related in distributing antenna system (Distributed AntennaSystem:DAS), based on L ₁a kind of resource management of norm constraint and the method for scheduling.

Background technology

Document [1] has proposed the concept of DAS structure first in 1987.In order to cover the covering blind spot of cellular cell, DAS system utilizes exposed coaxial cable to carry out with broadcasting signal.In DAS system, what antenna was all separated by distance is far away to reduce access distance, utilizes optical fiber, industrial siding or other radio frequency link to be connected with base station simultaneously.DAS system can also reduce the base station number in Serving cell, thereby has reduced operating cost, also more convenient maintenance.Research by scholar shows, DAS system all has obvious advantage aspect signal interference ratio, power and capacity ^{[2] [3]}.

Compared with traditional cell mobile communication systems, distributing antenna system has following features:

1) effectively improve cell coverage area:

DAS system is by being dispersed in antenna the covering completely that realizes community in community.This is also the original intention of framework DAS system.In traditional community, if user is positioned at cell edge, not only large scale fading ratio is larger, and has very large presence of intercell interference, makes overall SINR relatively low.And in DAS system, this situation be improved significantly.

2) antenna and user distance are short, decline little:

Because antenna is to be dispersed in community, therefore no matter in community where user, can find a wireless distributed module that distance oneself is very near.Distance is short just means that the power loss of signal in transmitting procedure is little, thereby has indirectly improved user's quality of wireless channel.

3) transmitting power diminishes, and interference diminishes:

Reason is wherein with above 2) say the same, due to apart from short, therefore, in order to reach identical service quality, DAS system just can have lower transmitting power than traditional community.So not only can improve the useful life of terminal use's battery, and can greatly reduce the interference between multi-user, capacity.

4) space diversity increases:

Due to all apart from each others of distributed module, therefore its spatial coherence is very little, like this with regard to Existential Space diversity gain.As long as make full use of obviously capacity of this character.

5) power efficiency is high:

In the spaced antenna structure of a simplification, the regional extent that little base station and 6 spaced antenna modules cover is

the radius of a circle of its homalographic is

in order to embody the fairness of comparison, suppose that the effective range that traditional base station covers is circular.Making the little base station of each DAS system and the power constraint of distributed module is P, and so total power is 7P.In traditional structure, suppose that path loss is L=d ^a, base station needs

power could realize identical coverage, wherein α be decline index.Therefore, the power efficiency of DAS structure is:

$\mathrm{\η}=\frac{{(12\sqrt{3}/\mathrm{\π}-1)}^{\frac{\mathrm{\α}}{2}}P}{7P}=\frac{{(12\sqrt{3}/\mathrm{\π}-1)}^{\frac{\mathrm{\α}}{2}}}{7}---(1-1)$

Therefore in the time of α=4, power efficiency gain η ≈ 6.54dB.So, identical overlay area, the needed power of DAS is lower.

There is at present scheduling of resource and the problem of management of a lot of documents in research DAS system.But its optimum target is all to merge (MRC:Maximal Ratio Combing) technology based on high specific.But in distributing antenna system research before, the merge algorithm of all supposing receiving terminal is desirable.Document [4] points out, along with merging the increase of way, the performance of the merge algorithms such as MRC is also just more and more responsive to the error of channel, in the time that way is greater than 8, even can reach the performance loss of 2dB.Therefore in real system, there is channel estimating equal error in receiving terminal, receiving terminal and transmitting terminal all can not obtain perfect channel information, the optimal performance that can obtain using merge algorithm as scheduling according to just obviously not obtaining optimum solution, also just cannot accomplish desirable MRC.The optimum target of document will there are differences with real system so above.

Summary of the invention

In order to overcome the deficiencies in the prior art, the present invention, to maximize system spectral efficiency as target, has proposed the distributing antenna system resource regulating method of a kind of combining adaptive sky line options; It approaches the optimal performance of merge algorithm and actual performance as far as possible.

The invention provides the distributing antenna system resource regulating method of a kind of combining adaptive sky line options, this algorithm ensureing under the prerequisite of throughput of system, taking the port that obtains rarefaction as target, taking the Power Limitation on each port as constraints.First it under the scene of distributing antenna system, utilize L1 norm to retrain the Power Limitation on each port, makes the port rarefaction of selection, thereby set up a kind of optimal model based on L1 norm constraint, obtains optimization equation; Then optimization equation is changed into dual problem and ask optimal solution, realize spaced antenna scheduling of resource.Concrete technical scheme is described below.

One, system model

In the system that the present invention considers, have M spaced antenna (DA:Distributed Antenna) port, each port has an antenna, has K any active ues and N subcarrier simultaneously.Make l _k[m, n] and h _k[m, n] represents respectively large scale decline and the impulse response of carrier wave n from port m to user k.

Suppose that at each scheduling moment channel be flat fading, and each subcarrier energy and can only distribute to a user.Definition ρ _k[n] is the indicator variable of a dynamic subcarrier assignment (SA:Sub-carrier Assignment).And if only if when subcarrier n distributes to user k, ρ _k[n]=1, otherwise ρ _k[n]=0.For the uniqueness that ensures to distribute, need to meet

suppose that base station end can obtain desirable channel condition information.

The total transmit power constraint of system is PT.Simple for what discuss, be averagely allocated to all ports herein.The power that each port can be got

this hypothesis can't reduce complexity discussed below.Definition p _k[m, n] is the transmitting power of port m in carrier wave n, in order to meet the Power Limitation (PCPP:Power Constraints Per Port) on each port:

$\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k\left[n\right]p_k[m,n]\≤P,p_k[m,n]\≥0,\∀m---(1-2)$

First suppose that all DA ports can transmit data on carrier wave arbitrarily.If the result calculating meets ρ _k[n]=1 and p _k[m, n]=0, shows that carrier wave n distributed to user k, but port m can not distribute any power to go the service into user k.This selection problem that just means port is also resolved naturally.

Receiving signal can be expressed as simultaneously:

y _k[n]=h _k[n]P _k[n]ω _k[n]x _k[n]+z _k[n],(1-3)

Wherein $h_k\left[n\right]=\left(\sqrt{l_k[1,n]}{h}_k\right[1,n],...,\sqrt{l_k[M,n]}{h}_k[M,n\left]\right),$ X _k[n] is the transmission symbol of user k, z _k[n] is that variance is σ ²additive Gaussian noise, ω _k[n] is weight coefficient,

the power distributing.In order to meet PCPP, optimum weight coefficient can be obtained by formula (1-4):

${\mathrm{\ω}}_k\left[n\right]=[\frac{h_k[1,n{]}^H}{\left|h_k\right[1,n\left]\right|},...,\frac{h_k{[M,n]}^H}{\left|h_k\right[M,n\left]\right|}{]}^T---(1-4)$

The signal to noise ratio receiving is so:

${\mathrm{\γ}}_k\left[n\right]=\frac{1}{{\mathrm{\σ}}^2}{\left(\underset{m=1}{\overset{M}{\mathrm{\Σ}}}\sqrt{G_k[m,n]p_k[m,n]}\right)}^2---(1-5)$

(1-6)

Wherein G _k[m, n]=l _k[m, n] | h _k[m, n] | ², power system capacity is C _k[n]=log ₂(1+ γ _k[n]).

In order to make the antenna port of selecting there is certain sparse property, therefore for this problem adds L1 norm constraint, represent the L1 norm constraint of port assignment power; .Obtain optimization problem (optimization equation) as follows:

$\min -\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k\left[n\right]C_k\left[n\right]+\mathrm{\δ}\left|\stackrel{\→}{p}\right|$

$\begin{array}{c}s.t.C1:\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k[m,n]\≤P,p_k[m,n]\≥0,\∀m;\\ C2:{\mathrm{\ρ}}_{k}n\∈\left\{\mathrm{0,1}\right\},\∀k,n,\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k\left[n\right]=1,\∀n\end{array}---(1-7)$

Two, optimum scheduling and resource are distributed

The present invention utilizes the method for Lagrange duality to carry out optimization to optimization problem; Comprise lagrange duality problem is resolved into N parallel subproblem, utilize geometry to try to achieve its closed solution; Utilize gradient descent method to solve Lagrangian coefficient.

Document [1] confirmed when subcarrier is enough large, and the duality gap of adaptive power division and dynamic allocation of carriers problem, close to zero, therefore can solve by lagrange duality problem herein.Its dual problem is:

$L(p,\mathrm{\ρ},\mathrm{\λ})=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k\left[n\right]C_k\left[n\right]-\mathrm{\δ}\left|\stackrel{\→}{P}\right|{\mathrm{\λ}}_j\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k\right[n\left]p_k\right[m,n]-P)---(1-8)$

Wherein λ=[λ ₁..., λ _m] be Lagrangian coefficient.In formula (8), constraint C2 does not take into account simultaneously, but in solution procedure below, can ensure that last solution meets this requirement.

The Lagrange duality form of formula (1-8) can be write as:

$g\left(\mathrm{\λ}\right)=\underset{P,\mathrm{\ρ}}{\max }(P,\mathrm{\ρ},\mathrm{\λ})---(1-9)$

Former problem can convert to:

$\begin{array}{cc}\underset{\mathrm{\λ}}{\min }& g\left(\mathrm{\λ}\right)\\ s.t.& {\mathrm{\λ}}_j\≤0,\∀j\end{array},---(1-10)$

In the time of fixing p and ρ, L (p, ρ, λ) is the linear equation about λ.And g (λ) is the optimal solution that maximizes these linear equations.Therefore the described optimization problem of formula (1-10) is a protruding optimization problem.

Lagrange duality equation is decoupled into the individual independently optimization problem of N here herein and solves, shown in (11):

$g\left(\mathrm{\λ}\right)=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{J}_n\left(\mathrm{\λ}\right)-\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\λ}}_{j}P---(1-11)$

Wherein:

$J_n\left(\mathrm{\λ}\right)=\underset{p,\mathrm{\ρ}}{\max }\left(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k\right[n\left]C_k\right[n]+\underset{j=1}{\overset{M}{\mathrm{\Σ}}}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\mathrm{\λ}}_j{\mathrm{\ρ}}_k[n\left]p_k\right[m,n]-\mathrm{\δ}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}|\sqrt{p_k[m,n]}\left|\right)---(1-12)$

2.1 sub-carrier power are distributed

Because each subcarrier can only be distributed to a user, therefore formula (1-12) has disclosed a rule of a dynamic assignment subcarrier, and that is exactly to be the specific user of each fixing sub-carrier selection, and formula (1-12) is maximized.When carrier wave n has been distributed to user k, so:

${\mathrm{\&rho;}}_j\left[n\right]=\left\{\begin{array}{cc}1& j=\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="1403171323221.png"\; state="\{\{state\}\}">k</figure-callout>}\\ 0& \mathrm{else}\end{array}{p}_k\right[m,n\left]\right\{\begin{array}{cc}\&GreaterEqual;0& j=k,\&ForAll;m\\ =0& j\&NotEqual;k,\&ForAll;m\end{array}---(1-13)$

Formula (1-13) substitution formula (1-12) can be obtained to following problem

$\underset{p}{\max }\left(C_k\right[n]+\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_j{p}_k[m,n]-\mathrm{\&delta;}\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}|\sqrt{p_k[m,n]}\left|\right)---(1-14)$

Formula (1-14) can obtain its optimal solution by following proof:

For convenience of description, special work stated as follows:

$q_m=\sqrt{-{\mathrm{\&lambda;}}_m}{p}_k[m,n],a_m=\frac{\sqrt{G_k[m,n]}}{\mathrm{\&sigma;}\sqrt{-{\mathrm{\&lambda;}}_m}},\&ForAll;m---(1-15)$

Formula (1-14) can become:

$\underset{q}{\max \log }(1+{\left(\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{a}_m{q}_m\right)}^2)-\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}(q_m^2+\mathrm{\&delta;}\frac{q_m}{\sqrt{-{\mathrm{\&lambda;}}_m}})---(1-16)$

After formula distortion, formula (1-16) becomes:

$\underset{q}{\mathrm{mqx}\log }(1+{\left(\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{a}_m{q}_m\right)}^2)-\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{(q_m+\frac{\mathrm{\&delta;}}{2\sqrt{-{\mathrm{\&lambda;}}_m}})}^2-\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&delta;}}^2}{{4\mathrm{\&lambda;}}_m}---(1-17)$

Order:

substitution formula (1-17) obtains:

$\underset{q}{\max }\log (1+{(\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{a}_m{r}_m-\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{a}_m\frac{\mathrm{\&delta;}}{2\sqrt{-{\mathrm{\&lambda;}}_m}})}^2)-\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{\left(r_m\right)}^2-\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&delta;}}^2}{{4\mathrm{\&lambda;}}_m}---(1-18)$

In the space of a M dimension, suppose vectorial a=(a ₁..., a _m) ^twith vectorial r=( _r1 ..., r _m) ^t, R represents point (r simultaneously ₁..., r _m).So for b>=0 arbitrarily, allly meet

r point formed a hyperplane { r|a ^tr=b}, vectorial a is its orthogonal vectors simultaneously.Be not difficult to find, for any R on hyperplane, all make a part above for formula (18) $\log (1+{({\mathrm{\&Sigma;}}_{m=1}^M{a}_m{r}_m-{\mathrm{\&Sigma;}}_{m=1}^M{a}_m(\mathrm{\&delta;}/2\sqrt{-{\mathrm{\&lambda;}}_m}))}^2)$ It is a definite value.

And for the Part II of formula (1-18), its expression be the distance of a R to the origin of coordinates, it should minimize.Can obviously find out all at hyperplane { r|a from Fig. 1 ^tpoint on r=b}, { focus of θ a| θ>=0} and hyperplane is nearest point to a directed quantity.That is to say in the optimal solution of formula (1-18), vectorial r should with vectorial a in the same way.That is to say that vectorial r and vectorial a are linear correlations.

Order: $\left[\begin{array}{ccc}d=\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{a}_m\frac{\mathrm{\&delta;}}{2\sqrt{-{\mathrm{\&lambda;}}_m}}& c=\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&delta;}}^2}{{4\mathrm{\&lambda;}}_m}{p}_m={\mathrm{ua}}_m& A=\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{a}_m^2\end{array}\right]$

Formula (1-18) can become:

$\underset{q}{\max \log }(1+{(\mathrm{uA}-d)}^2)-{\mathrm{Au}}^2-c---(1-19)$

Definition f (u)=log (1+ (uA+d) ²)-Au ²-c, formula (1-19) can equal zero and obtain its optimal solution by it is differentiated:

$\frac{\&PartialD;f\left(u\right)}{\&PartialD;u}|{u=u}^{*}=0---(1-20)$

After its differentiate, be a monobasic cubic function, can be by utilizing Bao He formula to ask.Bao Hewei asks simple cubic equation to propose a normalized form, supposes that its equation is x ³+ ax ²+ bx+c=0, its solution is:

$\left[\begin{array}{cc}{x}_1=\frac{(p+q)-2a}{6}& x_{\mathrm{2,3}}=-\left[\frac{(p+q)+4a}{12}\right]\&PlusMinus;i\left[\frac{\sqrt{3}(p-q)}{12}\right]---(1-21)\end{array}\right]$

Wherein $p={(m+n)}^{1/3},q={(m-n)}^{1/3},m=36\mathrm{ab}-{8a}^3-108c,n=[m^2+{(12b-{4a}^2)}^3{]}^{1/2}.$

2.2 dual problem gradients solve

Now can obtain the optimum power division under given λ, it is defined as p ^*(λ).By its substitution formula (1-12), we can obtain J _n(λ).Because g (λ) is a protruding optimization problem, therefore, in order to minimize dual problem, adopt gradient descent method herein.The gradient of g (λ) is:

${\mathrm{\&Delta;\&lambda;}}_j=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k^{*}\left[n\right]\left(\mathrm{\&lambda;}\right)-P,\&ForAll;j---(1-22)$

The iterative formula of so Lagrangian coefficient is provided by formula (1-23):

${\mathrm{\&lambda;}}_j^{l+1}=[{\mathrm{\&lambda;}}_j^l-S_{{\mathrm{\&lambda;}}_j}^l\left(\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k^{*}\right[n]\left(\mathrm{\&lambda;}\right)-P){]}^{+},\&ForAll;j---(1-23)$

The proof of formula (1-23):

Known according to formula (1-10):

$g\left({\mathrm{\&lambda;}}^{\&prime;}\right)=\underset{p,\mathrm{\&rho;}}{\max L}(p\left({\mathrm{\&lambda;}}^{\&prime;}\right),\mathrm{\&rho;}\left({\mathrm{\&lambda;}}^{\&prime;}\right),{\mathrm{\&lambda;}}^{\&prime;})---(1-24)$

Wherein:

$\begin{array}{c}L(p\left({\mathrm{\&lambda;}}^{\&prime;}\right),\mathrm{\&rho;}\left({\mathrm{\&lambda;}}^{\&prime;}\right),{\mathrm{\&lambda;}}^{\&prime;})=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k\left[n\right]\left({\mathrm{\&lambda;}}^{\&prime;}\right)C_k\left[n\right]\left({\mathrm{\&lambda;}}^{\&prime;}\right)-\mathrm{\&delta;}\left|\stackrel{\&RightArrow;}{P\left({\mathrm{\&lambda;}}^{\&prime;}\right)}\right|\\ +\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_j^{\&prime;}\left(\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k\right[n\left]\left({\mathrm{\&lambda;}}^{\&prime;}\right)p_k\right[m,n]\left({\mathrm{\&lambda;}}^{\&prime;}\right)-P)\end{array}---(1-25)$

Suppose ρ ^*[λ] and p ^*[λ] is the optimal solution that minimizes g (λ), so formula (1-26) set up.

$\begin{array}{c}g\left({\mathrm{\&lambda;}}^{\&prime;}\right)\&GreaterEqual;\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k^{*}\left[n\right]\left(\mathrm{\&lambda;}\right)C_k^{*}\left[n\right]\left(\mathrm{\&lambda;}\right)-\mathrm{\&delta;}\left|\stackrel{\&RightArrow;}{P^{*}}\left(\mathrm{\&lambda;}\right)\right|\\ +\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_j^{\&prime;}\left(\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k^{*}\right[n\left]\left(\mathrm{\&lambda;}\right)p_k^{*}\right[m,n]\left(\mathrm{\&lambda;}\right)-P)\end{array}---(1-26)$

And the right of formula (1-26) can be write as:

$g\left({\mathrm{\&lambda;}}^{\&prime;}\right)\&GreaterEqual;g\left(\mathrm{\&lambda;}\right)+\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}({\mathrm{\&lambda;}}_j^{\&prime;}-{\mathrm{\&lambda;}}_j)\left(\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k^{*}\right[n\left]\left(\mathrm{\&lambda;}\right)p_k^{*}\right[m,n]\left(\mathrm{\&lambda;}\right)-P)---(1-27)$

Beneficial effect of the present invention is: by technical scheme of the present invention, can naturally select from the antenna port close to targeted customer oneself service, and owing to there being the impact of L1 norm, its port obtaining will be sparse, the way that MRC merges like this can not be just very large (as being greater than 10), thereby not only can reduce the complexity of receiving terminal, and simulation model also approaches real wireless environment more.The most important thing is that its performance loss is very little.Also can find out by emulation below, algorithm has herein reduced the serve port quantity of selecting in the situation that ensureing systematic function, and the algorithm of this paper also has good convergence simultaneously.

Brief description of the drawings

Fig. 1 is the optimal direction of vectorial q.

Fig. 2 is the simulating area of different antennae port.

Fig. 3 is convergence.

Fig. 4 is the relation of system port number and the availability of frequency spectrum.

Fig. 5 is the relation that system port number and system are selected port sum.

Embodiment

Embodiment 1

If the parameter of embodiment:

Concrete implementation step:

Therefore L1-PG(L ₁-Proposed Greedy Algorithm) algorithm can be expressed as:

Simulation result:

Fig. 4 has shown the spectrum efficiency of the algorithm that the algorithm of L1 norm in this paper and document [6] propose and the graph of a relation of different antenna port number.Blue line is constant power algorithm, be exactly each antenna port by power uniform distribution the subcarrier to its each service.Along with the increase of antenna port, the spectrum efficiency of every kind of method all can be improved.Generally speaking, the lifting of its performance derives from antenna port increases and the diversity gain that brings.On the other hand, because antenna intersperses among in community, therefore it can reduce the distance of transmission.Therefore algorithm in this paper, with less with reference to the gap of algorithm, probably has 4% performance.But the performance of algorithm in this paper is still high a lot of than constant power algorithm.

Can know by upper surface analysis, can obtain in this paper based on L from Fig. 4 ₁the Performance Ratio of the optimized algorithm of norm constraint has 4% performance loss with reference to algorithm.But we can find by observing Fig. 5, under identical environment, the solution drawing by algorithm in this paper is than list of references, and total port number of its selection has 20% to 33% decline.Based on the description of text, we can know that sparse port selects, and it is very little can making the merge algorithm of actual system and the gap of desirable merge algorithm.The loss of 4% the theoretical upper limit that therefore Fig. 4 shows, can be less in real system, even may be the same.

Fig. 3 has shown convergence herein.Fig. 3 is illustrated with the example that is calculated as of λ.Generated at random four kinds of channels herein and realized, Fig. 3 has retouched out λ iterative value in these cases.Other variable in algorithm has similar convergence property.Fig. 3 has provided the F norm of λ error, can find out that algorithm in this paper all can be restrained in 10 iteration under any channel is realized, and therefore the convergence property of this paper is fine.

List of references

[1]A.A.M?Saleh,A.J.Rustako,and?R.S.Roman.Distributed?antennas?for?indoor?radio?communication[J].IEEE?Trans.Commn.Vol.35,pp.1245-1251,Dec,1987.

[2]M.V.Clark?et?al.Distributed?versus?centralized?antenna?arrays?in?broadband?wireless?networks[C].Proc?IEEE?Veb.Tecbnol.Conf.pp.33-37,Rhodes?Island,Greece,May2001.

[3]R.Hasegawa?et?al.Downlink?performance?of?a?CDMA?system?with?distributed?base?station.[C]Proc?IEEE?Veb.Tecbnol.Conf.pp.882-886,Oct2003.

[4]Tomiuk,Bohdan?R.A?new?look?at?maximal?ratio?combing.[C]GlobalTelecommunications?Conference,2000,GLOBERCOM’00.IEEE,vol.2pp.943-948,2000.

[5]W.Yu?and?R.Lui.Dual?methods?for?non?convex?spectrum?optimization?of?multicarrier?systems.Communications,[J]IEEE?Transaction?on.vol.54,no.7,pp.1310-1322,2006.

[6]R.Aggarwal,C.E.Koksal,and?P.Schniter.Joint?Scheduling?and?Resource?Allocation?in?OFDMA?Downlink?Systems?Via?ACK/NAK?Feedback[J].IEEE.Trans.Signal?Processing,vol.60,no.6,pp.3217-3227June2012。

Claims (3)

1. a distributing antenna system resource regulating method for combining adaptive sky line options, is characterized in that, first it under the scene of distributing antenna system, utilize L ₁norm retrains the Power Limitation on each port, makes the port rarefaction of selection, thereby sets up a kind of optimal model based on L1 norm constraint, obtains optimization equation; Then optimization equation is changed into dual problem and ask optimal solution, realize spaced antenna scheduling of resource.

2. distributing antenna system resource regulating method according to claim 1, is characterized in that: set up the method for the optimal model based on L1 norm constraint, specific as follows:

(1) determine in the system of considering and have M spaced antenna port, each port has an antenna, has K any active ues and N subcarrier simultaneously;

(2) determine the target function of optimization problem: in the situation that rarefaction port is selected, make the throughput of system reach maximum;

(3) determine the constraints of optimization problem: meet Power Limitation on each port and the uniqueness principle of allocation of carriers;

(4) set up optimization equation, shown in (1-7):

Wherein: ρ _k[n] is the indicator variable of a dynamic subcarrier assignment to user k; p _k[m, n] is the transmitting power of port m in carrier wave n;

for power system capacity; P represents the power that each port is got;

represent the L1 norm constraint of port assignment power.

3. distributing antenna system resource regulating method according to claim 1, is characterized in that: optimization equation is carried out to optimization, and idiographic flow is as follows:

(1) convert optimization equation shown in formula (1-7) to dual problem, shown in (1-8);

(2) dual problem is changed into and solve its lagrange duality problem, its form is suc as formula shown in (1-10);

Wherein:

(3) lagrange duality problem is decoupled into N independently duty Optimization, the optimization equation after its decoupling zero is converted into formula (1-12)

(4), to formula (1-12), solve the optimum power division p in given λ situation ^*(λ): utilize geometry to try to achieve its closed optimal solution;

(5) to formula (1-12), by p ^*(λ) substitution formula (1-12), and utilize gradient descent method to try to achieve its optimal solution p ^*(λ) and λ; Wherein:

ρ _k[n] is the indicator variable of a dynamic subcarrier assignment to user k; p _k[m, n] is the transmitting power of port m in carrier wave n;

for power system capacity; represent the L1 norm constraint of port assignment power; λ=[λ ₁..., λ _m] be Lagrangian coefficient; P represents the power that each port is got.

Images