Application Of Improved Genetic Algorithm In BS Placement Of The Wireless City Network

SCDMA (Synchronous Code Division Multiple Access) is the wireless accesssystem of independent information property rights. Daqing Oilfield communicationscompanies since 2002 is the first to adopt the technology. After launched wirelesstelephone "super PHS",with more than four years of vigorous development andconstruction,network has a certain size, basically covering the Daqing city, the centerof Daqing Oilfield area outside and the surrounding 4 counties, and the Anda City.asat present, the development of nearly 150,000 users. Subscription growth has beengoodmomentumofdevelopment.The same as other mobile communication system,such as GSM,CDMA,duringthe develop and constructive process of SCDMA wireless city network,programmingof the wireless source,especially the placement of the Base-station,is important forimprovingtheserviceofthewholenetworkandreducingthecostofproject construct.Nowit is verynecessarytoadopt awayforthewholeevaluationofSCDMAwirelesscity network,finding the unjustified problem in network designing,at the same timedirecttheprocesswellofthelaternetworkconstruction.Now,GSM,CDMA and etc. mobile communication's wireless network'researchemergeinendlessly.Mainlybasedonthe "simulatedannealing"ofthe planningstudy"Constrained for wireless network planning multi-layered targets", More research isthe use of "genetic algorithm" to the global optimization characteristics, Wirelessnetwork to solve multi-objective optimization problem (NP-complete). All of thesealgorithms for wireless network planning solution has played an important role inpromoting, But now the various algorithms in the algorithm itself, and specificapplications are still some shortcomings and deficiencies. For "simulated annealing",as the algorithm to process the withdrawal limit temperature conditions are harsh.Time Performance Optimization poor. For the "constrained multi-objectiveoptimization layered" algorithm, not only need to establish a different wirelessnetwork performance requirements of the "mathematical model" It also needs toproperlyhandletheconstraintsandtheobjectivefunction,isquitedifficult.Comparedto the above two algorithms, "genetic algorithm" by virtue of its problems is notdependent on the characteristics of global optimization, seem to solve the problem ofwireless network planning algorithm for the best. However, the current application tothe wireless network planning of the genetic algorithm, While genetic algorithmoptimization process "direction" and easily "premature" phenomenon, not improve,the large amount of calculation and easily into a "local optimal solution." otherexisting "genetic algorithm" for wireless network planning need to consider allaspectsofadifferentfocus,withoutamorecomplete algorithm to take full account of the practical application.This paper will fully understand the basic principles of genetic algorithms and"model theorem building block assumptions," such as mathematical theory on thebasis of from the optimization algorithm to accelerate the speed and avoid prematuretwo, the introduction of high quality and selection strategy, simplex crossoveroperator, fitness density variation found the median and climbing algorithm method,the genetic algorithm of the operation steps of improvement. And the improvedalgorithm applied to the Daqing Oilfield communications company SCDMA wirelesstelephone network base stations to distribution planning, in order to minimize theconstruction cost of inputs, a comprehensive upgrade of the entire integrated networkservice quality. Specific details are as follows :1. The definition of object functionAgainst SCDMA wireless telephone network area coverage and capacity ofbusiness coverage two key technical requirements, Base Station input costs andeconomic needs, the definition of wireless telephone network objective function :Where x indicated a specific base station distribution program (a specificsolution),is a vector{x1,x2,…xi}Area cover rate: f₁(x) =the area which receive strength bigger than limit/dispelspace area;Portfolio cover rate: f₂(x) =the sum of business weight which receive strengthbigger than limit/the whole sum of business weight;Base-station equipment cost: f₃(x) =(beforehand BS number upper limit-practicesetting BS number)/beforehand BS number upper limit;c (i-1,...,3) i is homologous evey cost function's weight,it depend on theintergrate require for network of operators.2, genetic algorithm improvements :(1) improvement of Selection operator:traditional roulette selection of the individual under the law fitness to determineits size selected probability, that is, the higher the fitness of individuals selected on thegreater probability; Conversely, the lower the fitness of individuals selectedprobability will become. But roulette law two shortcomings :roulette law can not guarantee high fitness individuals will be selected toenter the next generation population.Roulette law might adapt to the higher number of individuals elected to thenext generation population, but Gene has an excellent model to adapt tolower the individual to be selected as the next generation on the possibility ofgreatly reduced, so that can lead to a genetic algorithm deadly phenomenon,the emergence of precocious.This article will take a contemporary populations fitness highest individualmandatory Add the next generation groups, Removal of fitness as well as the highestindividual contemporary groups roulette choice. Thus, to achieve a high quality,effective control method, "precocious" phenomenon.(2) improvement of Cross operatortraditional genetic algorithms, cross-operator father behalf of the individual inthe implicit solution space for random value, Therefore no guarantee that theoperation of cross-offspring than individuals father substituting individual, amanifestation of blind search features, convergence for a long time.The traditional crossover operators generate new individual without directionalcharacteristics, the paper absorbed simplex thinking, the crossover operator isimproved. Use every generation the best individual cross-operation and participationof the arbitrary two individuals constitute a simplex, achieved the crossover of the twoindividuals on behalf of a father of the hybrid direction, thus increasing the optimalsolution to the direction of movement of probability, improve the speed optimizationmutation .(3) improvement of Mutation operatorthe traditional mutation operator is based on the specific engineering experiencefor the people, In the process of mutation if installed improperly, not only to increasethe diversity of population without any help, also increased the amount ofcomputation. In addition, the algorithm running in the various stages of mutationoperator magnitude of the demand is not necessarily the same.In addition to the mutation operator role, the paper while using the adaptivemutation operator, the evolutionary process of different stages of the mutationoperator value with the corresponding value of combining fitness, the diversity at thesame time guarantee that the genetic algorithm convergence. In addition, some fitnessdensity function:To contemporary individual to determine the distribution of, and then decided to adopta common gene mutation in the median or more mutations, Solutions to control thescope of the search space(4)Use "climbing algorithm to" solve "premature" phenomenonwhen GA forthcoming convergence. Most individuals have focused on theconvergence of the optimal solution or partial global optimal solution around. GAclimbing as weak, when the local population converges to the optimal solution, theindividual single gene model. through crossover and mutation operation has noteffectively increase the diversity of population, the introduction of the new entity, atthis time evolution almost to a halThis paper genetic algorithm "premature" phenomenon, especially usingclimbing algorithm. Assuming Stocks have occurred early convergence to localoptima L, L is the distance of around d Department there must be a higher degree ofadaptation to the local solution M, Local arbitrary close to the optimal solution L Nthe point opposite direction may exist point M, using expression:determine the coordinates of M, if the individual M fitness value of L greater than thefitness value, it serves as an individual M L offspring, Otherwise still prefer the Lpoint. Climbing algorithm to enhance the efficiency of local climbing ability, giventhe direction of evolution, promote algorithm jumping out of the local extreme point,Quick to find the global optimum.Through the above, the genetic algorithm improvements to enhance the entirealgorithm speed of convergence, effectively avoid the algorithm a "premature"phenomenon. I will improve the genetic algorithm with the SCDMA system withspecific parameters, on the computer and right-intensive 5km*5km City area targetedfor simulation, simulation results indicate that Convergence speed is the objectivefunction or requirement of the indicators have been effectively improved. In order toverify the validity of the algorithm, in the oilfields communications companies withthe relevant personnel, I used this paper to improve the genetic algorithm to fly inHsinchuang City SCDMA wireless networks for the overall assessment, also part ofthe base station distribution optimization, optimizing the wireless network structurehas become more rational, the whole network quality of service indicators improvedsignificantly. Verification test results show that this referred to the improved geneticalgorithms can be applied to fully SCDMA wireless network base stations actualdistribution planning projects.