Research On Neural Network Based On Genetic Algorithms For QoS Routing

With the rapidly progress and development of network technology ,there have been many new multimedia services in recent years, such as video on demand,distance learning and so on. These new service contributed greatly to the development of multicast communication, and also need to provide appropriate quality of service(QoS). Each user has different demand, such as allowed time for delay in data transmission,distortion in picture transmission or synchronous level in audio-visual. However, traditional routing technology only provide best-effort service,it can not ensure quality of service.As the explosive growth of voice and video business,QoS multicast routing has become a very important research subject in the areas of networks and distributed systems,and a variety of multicast mechanism satisfied the need of QoS also appeared in the areas of algorithms and protocol layers in recent years.In fact,the key of QoS problem is routing select in network,also known as path-finding problems,that means first set appropriate measure parameters as standards of way-finding, then start to select the path,and ensure that the data can be transmitted through the selected path form the source node to the destination node.In the past decade,neural network(NN) and genetic algorithms(GA) used very widely in protecting the quality of service.Both are abstract biological system of information computing and processing,the combination of both have caused the interest of the researchers.This article researched how to use the improved neural network algorithms to solve the QoS problems.This paper advanced a improved algorithm to solve the QoS problem of the bandwith,delay,delay jitter and the cost, mainly aimed at defect and shortage of NN,and combined with characteristic of GA for evolution.The paper introduced the QoS multicast routing problem,mainly discussed on its basic concepts and principles,math model and QoS metrics and so on,then analyzed the present condition and significance of the research on QoS multicast routing problem. This paper also described the basic principles of neural network and GA ,then analyzed the characteristic of this two algorithms and brought forward the shortcoming on the network quality of service.Then according to the characteristic of QoS multicast routing, combined with optimization features of GA,a new neural network based on genetic algorithm have been advanced. The new algorithm not only retained the strong global random search capability of genetic algorithm,but also has robustness and self-learning ability of neural networks,and also has extensive mapping capability of NN and rapid global convergence performance of Genetic algorithms. Genetic algorithms gived full play to its characteristics of global optimization in neural networks,but also can make up the shortage that responsed slowly and long traning time of neural networks.In the Matlab simulation environment,experimented on the improved algorithm and compared simulation results with existing schemes,then obtained that the algorithm indeed have a good performance and get the desired results on QoS problem by data analysis.