Evolutionary Learning Based Resource Allocation Algorithm For Wireless Communication Networks Of Xidian University

The resource allocation optimization problem in wireless communication networks is a hot research issue in the field of wireless communication. As kinds of new technologies appeared, the scarcity of spectrum resources increasing dramatically. Therefore, the research on the efficient allocation algorithm in wireless communication network has great practical significance to the country and the society. This thesis mainly focuses on solving the problem of low spectrum utilization ratio of existing spectrum allocation algorithm, and make great effort designing a variety of optimization models and algorithms for solving the problem of resource allocation in wireless communication networks Through the in-depth study of evolutionary algorithm, super heuristic technology, multi-objective optimization theory. The main works are as follows:Inspired from hyper-heuristic and graph theory, a graph-based hyper-heuristic algorithm(GHHA) is proposed to solve the spectrum allocation problem in multi-cell networks. The spectrum allocation problem in multi-cell networks is a NP-complete problem, which is very suitable to be solved by evolutionary algorithms. Searching pattern that directly searching for the optimal feasible solution by traditional evolutionary algorithm is broke when hyper-heuristic is turned up. A high level heuristic strategy is designed to search for the heuristic sequences with which the feasible solutions can be generated, and through this to obtain an optimal solution of the original optimization problem. The indirect searching method can mitigate the performance of operators and searching performance which are influenced by traditional direct searching method. GHHA construct an undirected graph based on spectrum allocation problem firstly, and then simulated annealing method is applied to search in the heuristic searching space which is consisting of low level heuristics to find a potential heuristic sequence which is corresponding to a optimal solution to the spectrum allocation problem. A variable neighborhood strategy is applied in the annealing process to effectively guide the searching. The algorithm is tested on a set of 20 benchmark problems, and the experimental results indicate that the overall performance of the proposed algorithm is relatively good especially on some difficult problems.A two-phase knowledge based hyper-heuristic algorithm(TPKHHA) is proposed for resource scheduling optimization problem in multi-cell networks through considering characteristics of the channel assignment problem and the related knowledge、rules obtained by solving it. The distribution of different demand matrices and the value of the value of interference matrix have great affect on the difficult when assigning to each cell, and inspired from the research results of the channel allocation algorithm based on graph symmetry, six prior knowledge based low level heuristics are presented. In the global searching process, harmony search algorithm is implemented to find potential heuristic sequence, and then simulated annealing strategy is continually executed on this sequence to search for the optimal assignment scheme in the local search process. Through the simulation experiment, the influence of each parameter value on the performance of TPKHHA is analyzed, and with the optimal parameter settings, 20 benchmark problems are tested. The experimental results show that the TPKHHA have competitive advantages compared with other classical spectrum allocation algorithms.A multi-objective joint routing and channel assignment model is constructed by optimizing system throughput and the cost simultaneously. Besides this, a multi-objective optimization algorithm(Improved NSGA-II Algorithm, INSGAII) is proposed to solve it. A novel local search operator is designed and a new population updating strategy is involved to overcome the useless of the crowding distance in original NSGA-II and a special phenomenon of the pareto set when solving the constructed multi-objective problem. In the decision-making step, a system robust measure is introduced to make the final selection among the pareto set. The simulation results show that the multi-objective modeling is a necessity and the proposed algorithm can obtain comparable pareto solutions in terms of diversity and convergence when dealing with the multi-objective route and channel assignment problem.