A kind of mobile communication Automatic Frequency Planning method that combines graph theory and genetic algorithm
Technical field
The present invention relates to moving communicating field, relate in particular to a kind of Automatic Frequency Planning method that combines the cellular mobile communication networks of graph theory and genetic algorithm.
Background technology
The design object of wireless network planning is to instruct engineering to be built into the accord with expectation traffic demand with minimum cost; Mobile communications network with certain grade of service mainly comprises the content of aspects such as information gathering, capacity planning, coverage planning, the prospecting of cloth station, frequency planning.Wherein, frequency planning is a very crucial technology, and the quality of frequency planning quality will play decisive influence to network quality.
At wireless communication field, frequency resource is precious resources.Improving the frequency spectrum resource utilization ratio promptly is in the limited frequency spectrum resources scope, under guaranteeing that network quality can received prerequisite, improves network capacity.Improve network capacity, just must reuse limited frequency resource.Though channeling has improved network capacity, but brought new problem---the deterioration of speech quality.Channeling is tight more, and the network interferences of bringing is also big more.The balance that how to obtain network capacity and speech quality is the problem that frequency planning must solve.That is to say that good frequency planning must be realized making full use of of frequency keeping on the basis of acceptable speech quality.
At present; Frequency planning method commonly used is to be bunch multiplex mode of representative with 4 * 3,3 * 3,2 * 6 etc. in the cellular mobile communication engineering; Though this kind frequency planning method is simple to operate; But because the real network construction is subject to terrain and its features, it is not the honeycomb shape of rule usually that the sub-district covers, so there are shortcomings such as locking frequency point, network interferences imbalance, planning low precision in the frequency planning result.
Chinese patent CN101047937A discloses a kind of mobile communication frequency planning method based on genetic algorithm; BCCH channel frequency point allocation scheme based on genetic algorithm has been proposed; But do not consider the frequency point allocation problem of TCH channel, GPRS channel and utilization frequency hopping, system frequency planning function that can only implementation part.
Chinese patent CN101331778A discloses a kind of equipment and method of cellular communication system being carried out frequency planning; Proposed to use interference matrix to carry out the method for frequency planning; But when number of cells increases; Interference matrix can increase with how much multiples, and arithmetic speed seriously descends, and can't be applicable to the cellular network of large scale deployment.
Summary of the invention
The objective of the invention is to the above-mentioned deficiency that exists to existing cell mobile communication systems, a kind of combination graph theory of Automatic Frequency Planning and the mobile communication Automatic Frequency Planning method of genetic algorithm fast and accurately of realizing is provided.
The present invention includes following steps:
1) obtains the received signal intensity that interior each sub-district, planning zone receives other sub-districts; Compare with signal strength threshold; Set up cell relations non-directed graph matrix, whether said cell relations non-directed graph matrix is used in the expression planning zone related between any two different districts;
2) select an interior sub-district of planning zone as current area;
3) according to said cell relations non-directed graph matrix, current area utilization greedy strategy is carried out the frequency point allocation first time, if distribute successfully, obtain first frequency of current area, then execution in step 2), all accomplish frequency point allocation for the first time up to all sub-districts;
If distribute failure, then reduce signal strength threshold, execution in step 1 then), rebulid cell relations non-directed graph matrix;
4) according to said cell relations non-directed graph matrix, the utilization graph-theoretical algorithm generates a plurality of frequency planning initial solutions, as the initial solution population of genetic algorithm;
5) the utilization genetic algorithm is carried out hereditary computing to said initial solution population, and until the end condition that satisfies genetic algorithm, the optimum individual of exporting current population is as the frequency planning result;
6) adopt said frequency planning result to carry out frequency point allocation to planning each sub-district in the zone.
In step 1), the said concrete steps of obtaining the reception signal strength signal intensity that each sub-district in the planning zone receives other sub-districts are following:
With sub-district k
jTransmitter place grid point is that the zone of lattice more than delimited at the center, and establishing the main plot is k
i, search said main plot k
iTo a plurality of reception signal strength signal intensities in said many lattice zone and average, this average is sub-district k
jReceive main plot k
iThe reception signal strength signal intensity.
The method to set up of said signal strength threshold is following:
Foundation includes the signal strength threshold array D of a plurality of threshold values, and the max-thresholds of getting said signal threshold value array D is signal strength threshold D
Th
The said concrete steps of setting up cell relations non-directed graph matrix are following:
More said each sub-district receives the reception signal strength signal intensity and the said signal strength threshold D of other sub-districts
Th, if the reception signal strength signal intensity that first sub-district receives second sub-district is greater than D
Th, think that then second sub-district is associated to first sub-district, and expression second sub-district in the cell relations non-directed graph matrix be designated as 1 to the corresponding positions that first sub-district is associated; Otherwise be designated as 0.
In step 3), the concrete steps that said utilization greedy strategy is carried out the frequency point allocation first time are following:
To each cell allocation frequency in the said planning zone, for each main plot k
i, search all and said main plot k
iIn the forbidding frequency is listed the frequency that said associated cell is used in related sub-district, and remaining available frequency constitutes available frequency set F
1With frequency and the adjacent forbidding frequency of frequently listing in thereof that said associated cell is used, remaining available frequency constitutes available frequency set F
2
If said available frequency set F
2Non-NULL is then searched F
2, the available frequency point allocation that numbering is minimum is given said main plot k
iIf F
2Be empty set, then search said available frequency set F
1, the available frequency point allocation that numbering is minimum is given said main plot k
i
Said if distribute failure, the concrete steps that then reduce signal strength threshold are following:
If said available frequency set F
1Be empty set, then be judged to be the Frequency Distribution failure, carry out the strategy of degenerating; Said degeneration strategy is that said current use signal strength threshold is deleted from said signal threshold value array, obtains new signal threshold value array, and the maximum signal threshold value of getting new signal threshold value array is threshold value D
Th
In step 4), said according to said cell relations non-directed graph matrix, the utilization graph-theoretical algorithm generates a plurality of frequency planning initial solutions, and is following as the concrete steps of the initial solution population of genetic algorithm:
Initial solution scale A is set;
For each the main plot k in the planning zone
i, search all and said main plot k
iIn the forbidding frequency is listed the frequency that said associated cell is used in related sub-district, and remaining available frequency constitutes available frequency set F
3With frequency and the adjacent forbidding frequency of frequently listing in thereof that said associated cell is used, remaining available frequency constitutes available frequency set F
4
If said frequency set F
4Non-NULL is then searched F
4, therefrom randomly draw an available frequency point allocation and give said current main plot k
iIf F
4Be empty set, then search said frequency set F
3, therefrom randomly draw an available frequency point allocation and give current main plot k
iIf F
3Also be empty set, then from all available frequencies that the user provides, randomly draw one and distribute to the main plot,, can obtain a frequency planning initial solution until distributing all sub-districts;
Repeat above operation, satisfy said initial solution scale A up to gained frequency planning initial solution quantity;
Said a plurality of frequency planning initial solution is as the initial solution population.
In step 5), said utilization genetic algorithm is following to the concrete steps that said initial solution population carries out hereditary computing:
(1) sets the genetic algorithm target function;
(2) with the initial solution population as contemporary population;
(3) calculate the ideal adaptation degree of contemporary population through the genetic algorithm target function, judge whether to meet the genetic algorithm end condition, then stop calculating if meet; Otherwise select, intersection, mutation operation obtain elementary progeny population, execution in step (4) then;
(4) take out 90% individuality from said elementary progeny population, merge as progeny population as contemporary population execution in step (3) with classic 10% individuality in the parent population.
In step (3), said algorithm end condition is following:
Ideal adaptation degree maximum surpasses the fitness preset value of target function in the contemporary population of said genetic algorithm, or;
Said genetic algorithm generation number surpasses 100, or;
Said genetic algorithm continuous 30 generation population the maximum of ideal adaptation degree do not change.
In step (3), the object function of said genetic algorithm is based on the same frequency interference value and the design of adjacent interference value frequently of a plurality of picked at random points in the planning zone.
The invention has the advantages that characteristics and demand according to the frequency planning of cell mobile communication systems; Graph theory and genetic algorithm are combined; Has the advantage that amount of calculation is little, accuracy is high; And can satisfy complicated requirements such as irregular terrain profiles landforms, large scale deployment, entire system frequency planning well, and solved the technical barrier that in complicated day by day wireless environment, carries out good frequency planning, satisfied fast-developing urban construction needs.
Description of drawings
Fig. 1 is the basic flow sheet of Automatic Frequency Planning.
Fig. 2 is the flow chart of graph-theoretical algorithm module.
Fig. 3 is the flow chart of genetic algorithm module.
Embodiment
Following examples will combine accompanying drawing that the present invention is done further elaboration.
Referring to Fig. 1, the basic procedure for Automatic Frequency Planning may further comprise the steps:
(1) imports the required initial information of frequency planning, like the division of grid sampled point, subdistrict position, antenna height or the like.
(2) function of graph-theoretical algorithm module is for genetic algorithm module generates a large amount of frequency planning initial solutions, comprises to set up non-directed graph, utilization greedy strategy, the main operation of three steps of generation initial solution population.
(3) a large amount of frequency planning initial solutions that graph-theoretical algorithm obtained import genetic algorithm module as the initial solution population and are optimized, to each for population select, intersection, mutation operation.
(4) according to the target function of plan objects design genetic algorithm, each is calculated ideal adaptation degree value for population substitution target function, if satisfy the end condition of genetic algorithm, the optimum individual of then exporting current population is as the frequency planning result.
Referring to Fig. 2, be the basic procedure of graph-theoretical algorithm module, below be detailed step:
(1) calculates the received signal intensity of interior each sub-district, planning zone to each grid point.
(2) set up signal strength threshold array D,, several threshold values are set by a fixed step size according to engineering experience;
(3) a plurality of threshold values of signal strength threshold array D, by ordering from big to small, the maximum of getting in the current array is signal strength threshold D
Th
(4) calculate reception signal strength signal intensity receive_cc (k between each sub-district
i, k
j).Concrete grammar is, with sub-district k
jTransmitter place grid point is that the zone of lattice more than delimited at the center, in the present embodiment, preferably adopts lattice zone, nine palaces, and establishing the main plot is k
i, search k
iReceive signal strength signal intensities and average to 9 of nine palace lattice zone, this receives signal average and is sub-district k
jReceive sub-district k
iSignal strength signal intensity, be designated as receive_cc (k
i, k
j).Note receive_cc (k
i, k
j) and receive_cc (k
j, k
i) unequal.
(5) relatively receive signal strength signal intensity receive cc (k
i, k
j) and signal strength threshold D
Th, set up cell relations non-directed graph matrix.This cell relations non-directed graph matrix is a two-dimensional matrix, if receive_cc (k
i, k
j) greater than D
Th, then think sub-district k
iCan be to sub-district k
jImpact, promptly these two sub-districts are relevant, with the k of cell relations non-directed graph matrix G
iRow k
jRow are designated as 1; Otherwise be designated as 0.
(6) calculation plot concerns that the number of each row " 1 " of non-directed graph matrix G obtains cell association degree G
i, press G
iSorted in the sub-district from big to small.
(7) be G with degree of association ordering
1The sub-district be starting point, the utilization greedy strategy, successively to the cell allocation frequency.Concrete grammar is, for each main plot k
i, search all and k
iRelated sub-district uses frequency to list the forbidding frequency in them, and remaining available frequency constitutes available frequency set F
1The frequency and the adjacent forbidding frequency of frequently listing in thereof that then they are used, remaining available frequency constitutes available frequency set F
2, F2 is the subclass of F1.
(8) if available frequency set F
2Non-NULL is then searched F
2, the available frequency point allocation that numbering is minimum is given current main plot k
iIf F
2Be empty set, then search available frequency and combine F
1, the available frequency point allocation that numbering is minimum is given current main plot k
i
(9) if available frequency set F
1It also is empty set; Then be judged to be and distribute failure; Carry out the strategy of degenerating, the signal strength threshold of current use is deleted from the signal strength threshold array, execution in step (3) is to (7) again; Until the first time frequency point allocation accomplish smoothly, explain that the cell relations non-directed graph matrix that uses this moment is only for the frequency planning of current network.
(10) initial solution scale A is set.
(11), use graph-theoretical algorithm to the cell allocation frequency according to cell relations non-directed graph matrix.Concrete grammar is, for each main plot k
i, search all and k
iIn the forbidding frequency is listed the frequency that they use in related sub-district, and remaining available frequency constitutes available frequency set F
3The frequency and the adjacent forbidding frequency of frequently listing in thereof that then they are used, remaining available frequency constitutes available frequency set F
4, F3 is the subclass of F4.
(12) if available frequency set F
4Non-NULL is then searched F
4, therefrom randomly draw an available frequency point allocation and give current main plot k
iIf F
4Be empty set, then search available frequency set F
3, therefrom randomly draw an available frequency point allocation and give current main plot k
iIf F
3Also be empty set, then from all available frequencies that the user provides, randomly draw one and distribute to this sub-district.
(13) be starting point with the G1 sub-district, to planning all sub-district execution in step (11) and (12) in the zone, obtain a frequency planning initial solution successively.
(14) repeating step (11) is to step (13), satisfies initial solution scale A up to the quantity of gained frequency planning initial solution.
Referring to Fig. 3, be the basic procedure of genetic algorithm module, below be detailed step:
(1) with the initial solution population of the frequency planning initial solution that obtains in the graph-theoretical algorithm as genetic algorithm.
(2) with the target function of the current population substitution genetic algorithm of genetic algorithm, calculate the ideal adaptation degree of current population, target function designs according to plan objects.
(3) judge whether the algorithmic end condition, if meet, the output optimized individual is as the algorithm optimal solution, and end is calculated; Otherwise turned to for the 4th step.
(4) carry out selection, intersection, the mutation operation of genetic algorithm according to certain probability, generate elementary filial generation.
(5) take out 90% individuality of elementary filial generation, merge as progeny population with classic 10% individuality in the parent population.Returned for the 2nd step.
Wherein, the algorithm end condition is:
(1) the ideal adaptation degree maximum in the current population surpasses the fitness preset value, and this preset value is determined by target function.
(2) the genetic algorithm generation number surpasses 100.
(3) continuous 30 generation population the maximum of ideal adaptation degree do not change, think that algorithm restrains.
The frequency of BCCH channel with the planning gsm system is an example below, and the design of genetic algorithm target function is described.
At first from all grid points of planning the zone, randomly draw X grid point, the same adjacent interference value frequently that calculates them is respectively CI
(i, j), CA
(i, j)
Add up following data successively:
The grid point number of writing sufficient 12dB≤CI≤22dB all over is numCI1;
The grid point number of satisfied-6dB≤CA≤6dB is numCA1;
The grid point number that satisfies CI<12dB is numCI2;
Satisfy CA<-the grid point number of 6dB is numCA2.
Again statistics is calculated as follows:
CI1(k)=numCI1/(numCI1+numCI2)
CA1(k)=numCA1/(numCA1+numCA2)
Last calculating target function:
SYD(k)=5×CI1(k)+5×CA1(k)
The fitness preset value is 10, and promptly the maximum of ideal adaptation degree reaches 10 in population, and then genetic algorithm stops.
For planning TCH channel, GPRS channel and the different demands such as planning under frequency-hopping mode thereof, a need carries out modify to the parameter of the target function of genetic algorithm and gets final product.