| Summary: | 碩士 === 國立屏東科技大學 === 資訊管理系 === 92 === To provide good QoS in large-scale communication networks, packets must be transmitted quickly and steadily. The goal of routing selection (RS) is to choose a set of routes that minimize the average delay of message. The cost of capacity and flow assignment (CFA) is composed of delay cost, fixed cost and variable cost. A smart and minimal cost of CFA not only save money but also improve the utility of networks. Although allocate low capacity to links can save fixed cost, it also increases delay cost. On the other side, although allocate high capacity to links can save delay cost, unfortunately, fixed cost may increase rapidly. Hence, RS and CFA are extremely complex network optimization problems.
Genetic Algorithm (GA) is an efficient and well-known heuristic optimization algorithm. But GA suffers premature that derive from the diversity of population is not enough. The long evolution time of GA is also needed. In this thesis, a Cost-Based Genetic Algorithm（CBGA）based on problem-space GA（PSGA） is proposed. The CBGA allocates resource of networks based on uniform distribution scheme, and it chooses the route and network capacity with the least cost firstly. Moreover, the CBGA can generate good initial population and evolve on the correct direction. Hence, it can decrease population size and generation size and thus obtain the optimal solution very quickly and steadily.
|
|---|
