CN105430656A - Load balancing method for super-dense heterogeneous mobile cellular network - Google Patents

Abstract

The invention provides a load balancing method for a super-dense heterogeneous mobile cellular network, comprising the steps of 1, dividing the coverage space of the super-dense heterogeneous mobile cellular network into a plurality of checkered grids with the same size, and periodically making statistics on the average flow rate in each checkered grid to obtain the loading capacity in each checkered grid, and using the loading capacity for representing time volatility and space difference of the network load; and 2, distributing base station providing coverage service corresponding to each checkered grid according to the loading capacity in each checkered grid and allocated wireless resources of each base station in the super-dense heterogeneous mobile cellular network. According to the load balancing method for the super-dense heterogeneous mobile cellular network, aiming at the characteristic that the loading capacities in 4G and post-4G mobile cellular networks are nonuniform in time variation and space distribution, effective load balancing is achieved, and utilization efficiency of wireless resources (such as fixed spectral bandwidths and fixed transmission efficiency of base stations) and system capacities in 4G and post-4G super-dense heterogeneous mobile cellular networks are improved.

Description

A kind of super-intensive isomery mobile cellular network load-balancing method

Technical field

The present invention relates to 4G and rear 4G mobile communication technology field, particularly relate to a kind of super-intensive isomery cellular network load-balancing method.

Background technology

Super-intensive isomery cellular network (Ultradensenetwork, UDN) owing to there is outstanding performance at capacity with improving in the network coverage, one of forth generation mobile communication and the 5th third generation mobile communication network networking mode Evolution Tendency is considered to.Its basic thought is a large amount of low power access points (AccessNode of hot zones dense deployment in macrocellular coverage, AN), as, microcell base station (microcellbasestation, and [HuaweiTechnologiesInc., the 5G:Atechnologyvision such as micromicro honeycomb base station (picocellbasestation, PcioBS) MicroBS), 2013, Whitepaper].But, this brings new challenge to traditional cellular network load balancing, this mainly due to, 1) the overlapping cellular network architecture covered makes low power base station effectively can not receive mobile subscriber, namely a large number of users is caused still to be attached to high power macrocell base stations, and its peripheral part mobile subscriber can only be received in small-power base station, unloading limited efficiency; 2) mobile network's load Distribution is subject to the impact of user distribution and mobile subscriber's different time data access amount, itself has time-varying characteristics and spatially distributed uneven characteristic.Therefore, effective load-balancing method is the necessary technology giving full play to super-intensive isomery beehive network system capacity boost ability and focus covering power.

Load-balancing technique based on cell breathing (cellbreathing) is the method that current industrial circle and academia mainly pay close attention to, its core concept, for: multiple abutting subdistrict is by the quantity of adjustment institute service-user, reaches the load balancing between multiple adjoining base station.The method realizing this thought can be divided into two classes: the direct mode controlling adjustment base station range with power, and with the indirect mode that cell edge or overlapping covered mobile subscriber low year base station reselection are selected.Document [Y.Bejerano, S.Han, CellBreathingTechniquesforLoadBalancinginWirelessLANs, IEEETransMobileComputing, 2009, transmitting power and then control WiFi access point footprint size by dynamically changing WiFi access point pilot signal 8 (6), pp.735-749] is proposed.For reducing energy consumption, document [L.Le, QoS-AwareBSSwitchingandCellZoomingDesignforOFDMAGreenCel lularNetworks, inProc.IEEEGlobecom2012, pp.1544-1549] devise power control algorithm based on Noncooperative game.Document [Z.Niu, et.Al, CellZoomingforCost-EfficientGreenCellularNetworks, IEEECommunicationsMagazine, 2010, pp.74-79] propose centralized and distributed cell breathing strategy, with two utility functions of its definition of optimization.Document [Y.Xu, H.Li, Z.Feng, etal., Energysustainabilitymodelingandliquidcellmanagementingre encellularnetworks, inProc.IEEEICC2013, pp.4414-4419], [H.Wang, H.Li, etal., LiquidCellManagementforReducingEnergyConsumptionExpenses inHybridEnergyPoweredCellularNetworks, inProc.IEEEWCNC2014, pp.1655-1660] have studied the impact of cell breathing on green energy resource energy supply cellular network energy utilization rate.

But existing method only considers mobile network's load capacity fluctuation in time, and cell breathing methods or in units of community or sector from the adjustment of cell edge center of housing estate coverage, or to select with Cell Edge User base station reselection.But, load capacity spatially obviously has otherness, and this Spatial Difference is effectively to embody this otherness in units of the annulus of unit community or sector, this has performance by this load-balancing method of reduction, such as, multiple cell edge load is all very low, and obvious direct mode and indirect mode all cannot reach the effect of load balancing.Therefore, improve resource utilization ratio by cell-breathing techniques and still have very large space, especially for super-intensive heterogeneous network, the overlap of a large amount of dissimilar base station is disposed and is made the topology of super-intensive heterogeneous network more complicated, and the Spatial Difference of load capacity is more complicated.

Summary of the invention

For the problems referred to above, the object of this invention is to provide a kind of new super-intensive isomery mobile cellular network load-balancing method.For load capacity in 4G and rear 4G mobile communications network time become and the uneven feature of spatial distribution, realize effective load balancing, improve Radio Resource in 4G and rear 4G super-intensive isomery mobile cellular network (e.g., fixed frequency spectrum bandwidth and base station fix transmitting power) utilization ratio and power system capacity.

For reaching above-mentioned purpose, the concrete technical scheme that the present invention takes is:

A kind of super-intensive isomery mobile cellular network load-balancing method, comprises the following steps:

S1, by several gridiron patterns equal sized by super-intensive isomery mobile cellular network coverage spatial division, and cycle statistics is carried out to the average discharge in each gridiron pattern, to obtain the load capacity in each gridiron pattern, and with the time fluctuation of its characterizing network load and Spatial Difference;

S2, according to the load capacity in each gridiron pattern, and the Radio Resource that in super-intensive isomery mobile cellular network, each base station configures, distributes the base station that each gridiron pattern correspondence covers service.

Wherein, step S1 comprises: be divided into super-intensive heterogeneous network overlay area | L| equal-sized gridiron pattern, and wherein, L is the set that all gridiron patterns are formed, | L| represents the quantity of this set interior element.

Defining each gridiron pattern is a load lattice TL (Traffic-lattices), makes X (k)=[x ₁(k), x ₂(k) ..., x _{| L|}(k)] ^trepresent that whole load lattice are in kth, the accumulative load capacity in k ∈ N number of cycle, wherein, k is different cycles mark, and N is Positive Integer Set, and x (k) represents the load capacity of certain gridiron pattern within a kth cycle, T representing matrix transposition symbol.

Wherein, tessellated size is selected according to load space resolution requirements.

Wherein, the described cycle duration according to statistic flow not in the same time correlation choose.The duration of a general measurement period is no less than 30 minutes.

Wherein, described Radio Resource comprises bandwidth sum power, and step S2 comprises:

By bandwidth and power decile, power be the unity emitter power P U (Powerunit) of ANn after W decile; Bandwidth be unit bandwidth after Hz decile, n is the mark of diverse access point AN, and the maximum transmission power of ANn and whole bandwidth are respectively P PU and P unit bandwidth.Wherein, P such as to represent at the quantity of point bandwidth sum power, is integer; N is the sign of access point AN, and such as, n=1 indicates AN1; Subscript n represents that base station indicates, 0 representation unit (without physical meaning);

According to AN current transmit power calculating ANn power consumption formula be: wherein, N _nfor the number of the transceiver of ANn, for the power consumption of ANn when minimum emissive power exports, ζ _nfor the power consumption parameter relevant with ANn load, determine according to different base station model.

According to formula calculate the average transmission rate R of mobile subscriber in TLi when all users in TLi are linked into ANn _i,n, wherein, i represents the mark of different TL, N ₀it is the standard deviation of the white noise of 0 for average; κ is path loss equation coefficient; M is path loss index, according to base station surrounding environment value between 2 to 4; d _i,nfor the maximum distance between ANn and TLi.Namely the distance access user farthest in TLi and the distance of ANn.

Definition ρ _i,nk whether () be assigned to the 0-1 variable of ANn for TLi in the kth cycle, is specifically defined as

In any period k, omit different cycles mark k, obtain optimized ρ _i,n(k) value

With represent and obtain all tessellated base stations that should distribute by that analogy by the base station n that gridiron pattern i should be assigned to; And accordingly base station is distributed.

obtain especially by following steps:

1) choice accuracy value ε >0; Calculate in the kth cycle, if TLi is associated with ANn, ANn needs the PU number consumed wherein, A represents the set of all access point AN, and L represents the set of all TL, R _i,nwith value; Definition t is iterations, and is initialized as t=0; Initializing variable $({\mathrm{\β}}_{i,n}^t,{\mathrm{\γ}}_n^t)\∈R,$ Make all associated variables ${\mathrm{\ρ}}_{i,n}^t=0,$ Initialization with be 0;

2) for each ANn, according to formula $M_n^t=\min \left\{\exp \left(\log \left(\frac{R_{i,n}}{{\mathrm{\Δ}}_n^0}\right)-{\mathrm{\χ}}_n^t+\frac{{\mathrm{\β}}_{i,n}^t-{\mathrm{\α}}_i^t}{U_{i,n}}\right),P\right\}$ Calculate wherein be the t time iteration, total transmitting power that ANn uses, by formula $M_n\left(k\right)=\underset{i\∈A}{\Σ}{\mathrm{\ρ}}_{i,n}\left(k\right)\·U_{i,n}\left(k\right),\∀n\∈A$ Calculate, wherein, P is the specified maximum transmission power of AN;

3) according to formula ${\mathrm{\θ}}^t=\exp (log\left({\mathrm{\Δ}}_n^0\right)-{\mathrm{\χ}}_n^t+\frac{{\mathrm{\β}}_{i,n}^t+{\mathrm{\α}}_i^t}{U_{i,n}})$ Calculate θ ^t;

4) according to P and θ ^tsize, to determine in the t time iteration optimum value, comprising:

If a. P≤θ ^t, determine according to following formula

${\mathrm{\&rho;}}_{i,n}^t=\left\{\begin{array}{cc}0,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}<0\\ \&lsqb;0,1\&rsqb;,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}=0\\ 1,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}>0\end{array}\right.,$

If b. P > θ ^t, determine according to following formula

${\mathrm{\&rho;}}_{i,n}^t=\left\{\begin{array}{cc}0,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t<0\\ \&lsqb;0,1\&rsqb;,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t=0\\ 1,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t>0\end{array}\right.;$

5) calculate according to following formula $({\mathrm{\&alpha;}}_i^t,{\mathrm{\&chi;}}_n^t)\&Element;N^{+}$ With $({\mathrm{\&beta;}}_{i,n}^t,{\mathrm{\&gamma;}}_n^t)\&Element;R:$

Wherein, middle expression variable or

6) according to formula ${\mathrm{\&kappa;}}_j^t={\mathrm{\&theta;}}^t\frac{D\left({\&CenterDot;}^t\right)-D_{est}^t}{\left|\right|\&part;D^{*}\left({\&CenterDot;}^t\right)|{|}^2},j\&Element;\left\{1,2,3,4\right\}$ Calculate wherein, $0<\underset{\&OverBar;}{\mathrm{\&theta;}}\&le;{\mathrm{\&theta;}}^t\&le;\stackrel{\&OverBar;}{\mathrm{\&theta;}}<2,$ θwith for scalar, for the estimated value of optimal solution, and according to formula calculate, wherein, when being the t time iterative computation, D ( ^t) optimal value;

7) according to formula ${\mathrm{\&alpha;}}_i^{t+1}={\mathrm{\&alpha;}}_i^t-{\mathrm{\&kappa;}}_1^t\&CenterDot;\&lsqb;\underset{n\&Element;A}{\&Sigma;}{\mathrm{\&rho;}}_{i,n}-1\&rsqb;,{\mathrm{\&beta;}}_{i,n}^{t+1}={\mathrm{\&beta;}}_{i,n}^t-{\mathrm{\&kappa;}}_2^t\&CenterDot;{\mathrm{\&rho;}}_{i,n},{\mathrm{\&gamma;}}_n^{t+1}={\mathrm{\&gamma;}}_n^t-{\mathrm{\&kappa;}}_3^t\&lsqb;P-M_n\&rsqb;,$ ${\mathrm{\&chi;}}_n^{t+1}={\mathrm{\&chi;}}_n^t-{\mathrm{\&kappa;}}_4^t\&lsqb;M_n-{\mathrm{\&Sigma;}}_{i=1}^{\left|L\right|}{\mathrm{\&rho;}}_{i,n}{U}_{i,n}\&rsqb;$ Calculate with

8) calculate D ( ^t+1), and calculate according to the following equation with

${\mathrm{\&phi;}}_j^{t+1}=\left\{\begin{array}{cc}{\mathrm{a\&phi;}}_j^t,& D\left({\&CenterDot;}^{t+1}\right)\&le;D^t\\ max\{{\mathrm{b\&phi;}}_j^t,c\},& D\left({\&CenterDot;}^{t+1}\right)>D^t\end{array}\right.$

9) t=t+1 is made;

10) calculating judges whether to meet stopping criterion for iteration: D ( ^t)-D ( ^t-1) < ε;

11) if above-mentioned stopping criterion for iteration meets, then stop iterative computation, meet if calculate end condition, then stop calculating, be optimal solution otherwise, perform step 2), continue iterative computation, until meet stopping criterion for iteration D ( ^t)-D ( ^t-1) till < ε.

Accompanying drawing explanation

Fig. 1 is method implementing procedure schematic diagram in embodiments of the invention;

Fig. 2 is super-intensive isomery mobile cellular network coverage spatially equal and opposite in direction gridiron pattern division schematic diagram in embodiments of the invention;

Fig. 3 characterizes schematic diagram based on the offered load amount time variation of load lattice and Spatial Difference in embodiments of the invention.

Embodiment

For making above-mentioned feature and advantage of the present invention become apparent, special embodiment below, and coordinate institute's accompanying drawing to be described in detail below.

The technical problem to be solved in the present invention is: for the feature become during load capacity in mobile communications network and spatial distribution is uneven, how to design a kind of effective load-balancing method, improve Radio Resource in 4G and rear 4G super-intensive isomery cellular network (e.g., fixed frequency spectrum bandwidth and base station fix transmitting power) utilization ratio and power system capacity.

For solving the problems of the technologies described above, the invention provides a kind of super-intensive isomery mobile cellular network load-balancing method, as shown in Figure 1, comprising the following steps:

S1, certain super-intensive isomery mobile cellular network coverage is spatially divided into equal-sized gridiron pattern, and cycle statistics is carried out to the average discharge in gridiron pattern, as shown in Figure 3, time fluctuation and the Spatial Difference of offered load is levied with different gridiron pattern internal burden scale, as can be seen from Figure 3, the load capacity difference (i.e. color) of same period (i.e. time), different gridiron pattern (i.e. space); The load capacity of different cycles (i.e. time), same gridiron pattern (i.e. space) also different (i.e. color).

S2, carry out cycle statistical value according to the average discharge in each gridiron pattern, and the Radio Resource that each base station of super-intensive isomery mobile cellular network configures, the base station that different gridiron pattern correspondence covers service is distributed.

Wherein, step S1 is specially: be divided into as shown in Figure 2 by certain super-intensive heterogeneous network overlay area | the chessboard of L| equal-sized gridiron pattern composition, wherein, L is the set that all gridiron patterns are formed, | L| represents the quantity of this set interior element, is set by tessellated size, can select in engineering practice according to load space resolution requirements, such as, 5 meters, 10 meters, 15 meters etc. are chosen as.Claim each gridiron pattern to be a load lattice (Traffic-lattices, TL), make X (k)=[x ₁(k), x ₂(k) ..., x _{| L|}(k)] ^trepresent that whole load lattice are in kth, the accumulative traffic carrying capacity in k ∈ N number of cycle, wherein, k is different cycles mark, and N is Positive Integer Set.

Wherein, step S2 is specially:

By bandwidth and power two Radio Resource decile, w and hz is respectively unity emitter power (Powerunit, PU) and the unit bandwidth of ANn after decile, and n is that different AN identifies, and in engineering practice, they can decile as required, e.g., and unity emitter power 1mW, unit bandwidth 12kHz.The maximum transmission power of ANn and whole bandwidth are respectively P PU and P unit bandwidth.

Calculating ANn power consumption can according to AN current transmit power according to formula calculate, wherein, N _nfor the number of the transceiver of ANn, for the power consumption of ANn when minimum emissive power exports, ζ _nfor the power consumption parameter relevant with ANn load, can determine according to different base station model in engineering practice.

According to formula calculate the average transmission rate of mobile subscriber in TLi when all users in TLi are linked into ANn, wherein, N ₀it is the standard deviation of the white noise of 0 for average; κ is path loss equation coefficient, can obtain from communication engineering handbook; M is path loss index, and it is according to base station surrounding environment value between 2 to 4; d _i,nfor the maximum distance between ANn and TLi.

Definition ρ _i,nk whether () be assigned to the 0-1 variable of ANn for TLi in the kth cycle, has and be defined as in any period k, omit different cycles mark k, optimized ρ _i,n(k) value obtain especially by following steps:

1) choice accuracy value ε >0, such as value 0.001; Calculate (in the kth cycle, if TLi is associated with ANn, ANn needs the PU number consumed), R _i,nwith value; Make t be iterations, and be initialized as t=0; Initializing variable make all associated variables ${\mathrm{\&rho;}}_{i,n}^t=0,$ Initialization with be 0;

2) for each ANn, according to formula $M_n^t=\min \left\{\exp \left(\log \left(\frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}\right)-{\mathrm{\&chi;}}_n^t+\frac{{\mathrm{\&beta;}}_{i,n}^t-{\mathrm{\&alpha;}}_i^t}{U_{i,n}}\right),P\right\}$ Calculate wherein be the t time iteration, total transmitting power that ANn uses, according to formula $M_n\left(k\right)=\underset{i\&Element;A}{\&Sigma;}{\mathrm{\&rho;}}_{i,n}\left(k\right)\&CenterDot;U_{i,n}\left(k\right),\&ForAll;n\&Element;A$ Calculate, P is the specified maximum transmission power of AN;

3) according to formula ${\mathrm{\&theta;}}^t=\exp \left(log\right({\mathrm{\&Delta;}}_n^0)-{\mathrm{\&chi;}}_n^t+\frac{{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t}{U_{i,n}})$ Calculate θ ^t;

4) according to P and θ ^tsize, to determine in the t time iteration optimum value:

If c. P≤θ ^t, determine according to following formula

${\mathrm{\&rho;}}_{i,n}^t=\left\{\begin{array}{cc}0,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}<0\\ \&lsqb;0,1\&rsqb;,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}=0\\ 1,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}>0\end{array}\right.,$

If d. P> θ ^t, determine according to following formula

${\mathrm{\&rho;}}_{i,n}^t=\left\{\begin{array}{cc}0,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t<0\\ \&lsqb;0,1\&rsqb;,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t=0\\ 1,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t>0\end{array}\right.$

5) calculate according to following formula with

Wherein, middle expression variable or

6) according to formula ${\mathrm{\&kappa;}}_j^t={\mathrm{\&theta;}}^t\frac{D\left({\&CenterDot;}^t\right)-D_{est}^t}{\left|\right|\&part;D^{*}\left({\&CenterDot;}^t\right)|{|}^2},j\&Element;\left\{1,2,3,4\right\}$ Calculate wherein, $0<\underset{\&OverBar;}{\mathrm{\&theta;}}\&le;{\mathrm{\&theta;}}^t\&le;\stackrel{\&OverBar;}{\mathrm{\&theta;}}<2,$ θwith for scalar, for the estimated value of optimal solution, and according to formula calculate, wherein, when being the t time iterative computation, D ( ^t) optimal value.

7) according to formula ${\mathrm{\&alpha;}}_i^{t+1}={\mathrm{\&alpha;}}_i^t-{\mathrm{\&kappa;}}_1^t\&CenterDot;\&lsqb;\underset{n\&Element;A}{\&Sigma;}{\mathrm{\&rho;}}_{i,n}-1\&rsqb;,{\mathrm{\&beta;}}_{i,n}^{t+1}={\mathrm{\&beta;}}_{i,n}^t-{\mathrm{\&kappa;}}_2^t\&CenterDot;{\mathrm{\&rho;}}_{i,n},{\mathrm{\&gamma;}}_n^{t+1}={\mathrm{\&gamma;}}_n^t-{\mathrm{\&kappa;}}_3^t\&lsqb;P-M_n\&rsqb;,$ ${\mathrm{\&chi;}}_n^{t+1}={\mathrm{\&chi;}}_n^t-{\mathrm{\&kappa;}}_4^t\&lsqb;M_n-{\&Sigma;}_{i=1}^{\left|L\right|}{\mathrm{\&rho;}}_{i,n}{U}_{i,n}\&rsqb;$ Calculate with

8) calculate D ( ^t+1), and calculate according to the following equation with

${\mathrm{\&phi;}}_j^{t+1}=\left\{\begin{array}{cc}{\mathrm{a\&phi;}}_j^t,& D\left({\&CenterDot;}^{t+1}\right)\&le;D^t\\ max\{{\mathrm{b\&phi;}}_j^t,c\},& D\left({\&CenterDot;}^{t+1}\right)>D^t\end{array}\right.$

9) t=t+1 is made;

10) calculating judges whether to meet stopping criterion for iteration: D ( ^t)-D ( ^t-1) < ε;

11) if above-mentioned stopping criterion for iteration meets, then stop iterative computation, meet if calculate end condition, then stop calculating, be optimal solution otherwise, perform step 2), continue iterative computation, until meet stopping criterion for iteration D ( ^t)-D ( ^t-1) till < ε.

For load capacity in mobile communications network time become and the uneven feature of spatial distribution, design one super-intensive isomery mobile cellular network load-balancing method, levies time fluctuation and the Spatial Difference of offered load amount with different gridiron pattern internal burden scale; The base station different gridiron pattern correspondence being covered to service is distributed, and greatly improves 4G and Radio Resource in rear 4G super-intensive isomery cellular network (e.g., fixed frequency spectrum bandwidth and base station fix transmitting power) utilization ratio and power system capacity.

Claims (7)

1. a super-intensive isomery mobile cellular network load-balancing method, comprises the following steps:

S1, by several gridiron patterns equal sized by super-intensive isomery mobile cellular network coverage spatial division, and cycle statistics is carried out to the average discharge in each gridiron pattern, to obtain the load capacity in each gridiron pattern, and with the time fluctuation of its characterizing network load and Spatial Difference;

S2, according to the load capacity in each gridiron pattern, and the Radio Resource that in super-intensive isomery mobile cellular network, each base station configures, distributes the base station that each gridiron pattern correspondence covers service.

2. method according to claim 1, is characterized in that, step S1 comprises: be divided into super-intensive heterogeneous network overlay area | L| equal-sized gridiron pattern, and wherein, L is the set that all gridiron patterns are formed, | L| represents the quantity of this set interior element;

Defining each gridiron pattern is a load lattice TL, makes X (k)=[x ₁(k), x ₂(k) ..., x _{| L|}(k)] ^trepresent that whole load lattice are in kth, the accumulative load capacity in k ∈ N number of cycle, wherein, k is different cycles mark, and N is Positive Integer Set, and x (k) represents the load capacity of certain gridiron pattern within a kth cycle, T representing matrix transposition symbol.

3. method according to claim 1, is characterized in that, described tessellated size is selected according to load space resolution requirements.

4. method according to claim 1, is characterized in that, the duration in described cycle according to statistic flow not in the same time correlation choose.

5. method according to claim 1, is characterized in that, described Radio Resource comprises bandwidth sum power.

6. method according to claim 5, is characterized in that, step S2 comprises:

By bandwidth and power decile, power be the unity emitter power P U of ANn after decile; Bandwidth it is unit bandwidth after decile; Wherein n is the mark of diverse access point AN, and subscript n represents that base station indicates, 0 representation unit;

According to AN current transmit power calculating ANn power consumption formula be: wherein, N _nfor the number of the transceiver of ANn, for the power consumption of ANn when minimum emissive power exports, ζ _nfor the power consumption parameter relevant with ANn load;

According to formula calculate the average transmission rate R of mobile subscriber in TLi when all users in TLi are linked into ANn _i,n, wherein, i represents the mark of different TL, N ₀it is the standard deviation of the white noise of 0 for average; κ is path loss equation coefficient; M is path loss index; d _i,nfor the maximum distance between ANn and TLi;

Definition ρ _i,nk whether () be assigned to the 0-1 variable of ANn for TLi in the kth cycle, is specifically defined as

In any period k, omit different cycles mark k, obtain optimized ρ _i,n(k) value

With represent and obtain all tessellated base stations that should distribute by that analogy by the base station n that gridiron pattern i should be assigned to; And accordingly base station is distributed.

7. method according to claim 6, is characterized in that, obtain especially by following steps:

1) choice accuracy value ε >0; Calculate in the kth cycle, if TLi is associated with ANn, ANn needs the PU number consumed wherein, A represents the set of all access point AN, and L represents the set of all TL, R _i,nwith value; Definition t is iterations, and is initialized as t=0; Initializing variable make all associated variables initialization with be 0;

2) for each ANn, according to formula $M_n^t=min\{\exp \left(\log \right(\frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0})-{\mathrm{\&chi;}}_n^t+\frac{{\mathrm{\&beta;}}_{i,n}^t-{\mathrm{\&alpha;}}_i^t}{U_{i,n}}),P\}$ Calculate wherein be the t time iteration, total transmitting power that ANn uses, by formula $M_n\left(k\right)=\underset{i\&Element;A}{\&Sigma;}{\mathrm{\&rho;}}_{i,n}\left(k\right)\&CenterDot;U_{i,n}\left(k\right),\&ForAll;n\&Element;A$ Calculate, wherein, P is the specified maximum transmission power of AN;

3) according to formula ${\mathrm{\&theta;}}^t=\exp \left(\log \right({\mathrm{\&Delta;}}_n^0)-{\mathrm{\&chi;}}_n^t+\frac{{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t}{U_{i,n}})$ Calculate θ ^t;

4) according to P and θ ^tsize, to determine in the t time iteration optimum value, comprising:

If a. P≤θ ^t, determine according to following formula

${\mathrm{\&rho;}}_{i,n}^t=\left\{\begin{array}{cc}0,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}<0\\ \&lsqb;0,1\&rsqb;,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}=0\\ 1,& {\mathrm{\&beta;}}_{i,n}^t+U_{i,n}\log \frac{R_{i,n}}{{\mathrm{\&Delta;}}_n^0}>0\end{array}\right.,$

If b. P > θ ^t, determine according to following formula

${\mathrm{\&rho;}}_{i,n}^t=\left\{\begin{array}{cc}0,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t<0\\ \&lsqb;0,1\&rsqb;,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t=0\\ 1,& U_{i,n}{\mathrm{\&chi;}}_n^t-{\mathrm{\&beta;}}_{i,n}^t+{\mathrm{\&alpha;}}_i^t>0\end{array}\right.;$

5) calculate according to following formula $({\mathrm{\&alpha;}}_i^t,{\mathrm{\&chi;}}_n^t)\&Element;N^{+}$ With $({\mathrm{\&beta;}}_{i,n}^t,{\mathrm{\&gamma;}}_n^t)\&Element;R:$

Wherein, middle expression variable or

6) according to formula ${\mathrm{\&kappa;}}_j^t={\mathrm{\&theta;}}^t\frac{D\left({\&CenterDot;}^t\right)-D_{est}^t}{\left|\right|\&part;D^{*}\left({\&CenterDot;}^t\right)|{|}^2},j\&Element;\{1,2,3,4\}$ Calculate wherein, $0<\underset{\&OverBar;}{\mathrm{\&theta;}}\&le;{\mathrm{\&theta;}}^t\&le;\stackrel{\&OverBar;}{\mathrm{\&theta;}}<2,$ θwith for scalar, for the estimated value of optimal solution, and according to formula calculate, wherein, when being the t time iterative computation, D ( ^t) optimal value;

7) according to formula ${\mathrm{\&alpha;}}_i^{t+1}={\mathrm{\&alpha;}}_i^t-{\mathrm{\&kappa;}}_1^t\&CenterDot;\&lsqb;\underset{n\&Element;A}{\&Sigma;}{\mathrm{\&rho;}}_{i,n}-1\&rsqb;,{\mathrm{\&beta;}}_{i,n}^{t+1}={\mathrm{\&beta;}}_{i,n}^t-{\mathrm{\&kappa;}}_2^t\&CenterDot;{\mathrm{\&rho;}}_{i,n},{\mathrm{\&gamma;}}_n^{t+1}={\mathrm{\&gamma;}}_n^t-{\mathrm{\&kappa;}}_3^t\&lsqb;P-M_n\&rsqb;,$ ${\mathrm{\&chi;}}_n^{t+1}={\mathrm{\&chi;}}_n^t-{\mathrm{\&kappa;}}_4^t\&lsqb;M_n-{\mathrm{\&Sigma;}}_{i=1}^{\left|L\right|}{\mathrm{\&rho;}}_{i,n}{U}_{i,n}\&rsqb;$ Calculate with

8) calculate D ( ^t+1), and calculate according to the following equation with

${\mathrm{\&phi;}}_j^{t+1}=\left\{\begin{array}{cc}{\mathrm{a\&phi;}}_j^t,& D\left({\&CenterDot;}^{t+1}\right)\&le;D^t\\ max\{{\mathrm{b\&phi;}}_j^t,c\},& D\left({\&CenterDot;}^{t+1}\right)>D^t\end{array}\right.$

9) t=t+1 is made;

10) calculating judges whether to meet stopping criterion for iteration: D ( ^t)-D ( ^t-1) < ε;

11) if above-mentioned stopping criterion for iteration meets, then stop iterative computation, meet if calculate end condition, then stop calculating, be optimal solution otherwise, perform step 2), continue iterative computation, until meet stopping criterion for iteration D ( ^t)-D ( ^t-1) till < ε.