一個以分類元系統建構之動態適應式網路路徑決定方法

Abstract

近年來網路快速地發展，有越來越多的電腦連上網路，因此網路的連結成為一項非常重要的技術。為了能使網路順利運轉，需要一套有效率的routing protocol。由於routing protocol是以分散式的方式，由每台路由器各自執行，經過一段時間的彼此溝通後，建立起各自的routing table。因此routing protocol須要仔細地設計，才能正確無誤的運作。 本論文的研究目標，是嘗試藉由演化式計算的精神，將classifier system應用在網路的routing問題上：在每個網路連結節點上放置一個classifier system，藉得其自我調適（self-adaptive）的特性，來適應變動的網路環境，達到傳統路由器與routing protocol的功能，建構出該節點之routing table，並能隨著網路流量的改變，而動態修改所選擇之路徑。最後的實驗結果，則證明了本系統的良好表現。

In recent years, the network develops rapidly, and more and more computers are connected with it. Hence, interconnection of networks became a very import technology. In order to make networks work properly, an effective routing protocol is needed. Because routers run the routing protocol in a distributed way, a routing protocol has to be designed very carefully to work correctly. The purpose of this thesis is to apply the evolutionary computation to the routing problem. We place a classifier system on each router. These classifier systems will be able to achieve the task of traditional routing protocols—establish the routing table—after training. In addition, they will adapt to the changing network environment and choose the best path accordingly. The final results of experiments showed the good performance of this system.