Studies On Channel Assignment Problem Based On Genetic Algorithm And Particle Swarm Optimization Algorithm

As the demand for mobile telecommunication grows, how to reuse available frequency in different cells without violation of interferences between mobile stations to meet the cell's call requirement becomes more and more important for improvement of system capacity and efficient use of radio spectrum.An improved genetic algorithm (GA) with low computation complexity is proposed to solve the existed drawbacks on low convergence speed and weak capacity of local search in traditional GA based solutions for CAP. The channel assignment matrix is encoded by a minimum-separation encoding scheme, which fulfills the traffic requirement and co-site constrains inherently and reduces the solution space. The initial approach of max demand first with least conflicts firstly assigns channels to the cells with the largest channel requirement, which reduces the conflicts and improves the quality of initial solution. In order to keep the diversity of population, the selective mutate operator with local search can only select the elements, which is helpful for the improvement of individual's fitness to mutate.Although the improved GA based CAP scheme has a faster convergence speed, the convergence rate of optimization is not ideal. Particle swarm optimization (PSO) algorithms do not have the same complex crossover and mutate operations with GA. Therefore, the computation complexity is reduced greatly and the speed of convergence is faster for the parallel computation of particles. The easy-realized PSO has less parameter, which can be adjusted conveniently. The PSO has been applied to solve the CAP and works well.Further, a discrete particle swarm optimization (DPSO) and frequency exhaust assignment (FEA) combined scheme for CAP is proposed to solve the existed drawbacks on low convergence speed and weak capacity of local search in current DPSO algorithms. The motion equation is defined by experimental research on inertia selection and parameters adjustment to produce the best call orderings, which is assigned by FEA to meet the large and unevenly distributed traffic demand for most cells. The result shows that the proposed channel assignment scheme obtains the convergence rate of 100% for all benchmark problems with a fast speed. The performance does not depend on the quality of initial solution, which proves an efficient and stable approach.