In traditional networks, routing table is essential for packet transmission due to the lack of the direction information abut destination in the head of packet. However, it is feasible to make the address of device encode the routing information with the application of data technology. In this article, we propose new identities for networking routers –vectors, and a new routing principle based on these vectors is designed accordingly. These vectors encode the device distance information and serve as a pattern of the network topology. Then, routing decisions could be made by these vector calculations and only requirement of table query on the destination vector following the proposed routing principle. The proposed method is not limited in calculating the shortest path routing, but extend to solve the constrain routing problem. Besides, multi-paths routing is also available as long as multi-paths exist between the origin-destination pairs. The simulation results show that our proposed method works reliable and stable in routing tasks, and can achieve a remarkable performance when compared with the state-of-the-art work on the delay constrained least cost path (DCLC) problem.

中文翻译：

---

基于学习的域内 QoS 路由方法

在传统网络中，由于数据包头部缺少与目的地相关的方向信息，因此路由表对于数据包传输至关重要。但是，利用数据技术使设备地址编码路由信息是可行的。在本文中，我们为网络路由器提出了新的身份——向量，并据此设计了基于这些向量的新路由原则。这些向量对设备距离信息进行编码，并用作网络拓扑的模式。然后，可以通过这些向量计算做出路由决策，并且仅需要遵循所提出的路由原则对目标向量进行表查询。所提出的方法不限于计算最短路径路由，而是扩展到解决约束路由问题。除了，只要源-目的地对之间存在多路径，也可以使用多路径路由。仿真结果表明，我们提出的方法在路由任务中工作可靠且稳定，与延迟约束最小成本路径（DCLC）问题的最新工作相比，可以取得显着的性能。