一种智能多路径路由及子流分配协同算法

传统单一路径的传输机制难以满足当前以智慧城市为代表的新一代应用对时延、丢包率等网络性能的要求，而现有多路径传输机制在路由算法及子流分配等方面不能根据网络实时状态调整且互相缺乏协同。引入强化学习理论并结合软件定义网络，提出多路径路由及子流分配协同算法。基于Q-learning设计多路径路由算法，并从策略协同角度对其进行改进，实现路由与子流分配的相互协同。在此基础上，通过Q-value的回环消除方法保证路由准确性并提高算法收敛速度。实验结果表明，该算法在网络负载动态变化过程中能实时调整最佳的多路径路由及子流分配协同策略，提高了传输成功率。

An Intelligent Cooperative Algorithm for Multi-Path Routing and Subflow Allocation

The new generation of applications such as smart cities put forward higher requirements for network performance, including lower delay and packet loss rate.However, the traditional single-path transmission mechanism fail to meet the new requirements, while the existing multi-path transmission technology is limited by some problems in routing and subflow allocation, such as the difficulty in adjusting strategies according to the real-time network conditions and the lack of cooperation.An intelligent cooperative algorithm for multi-path routing and subflow allocation is proposed, which is based on the reinforcement learning theory and software defined network.For this algorithm, a multi-path routing method based on Q-learning is designed and improved from the perspective of strategy coordination to realize mutual coordination between routing and subflow allocation. On this basis, a loop elimination method based on Q-value is used to ensure the validity of routing decision and to improve the convergence speed of the algorithm.The experimental results show that the proposed algorithm can adjust the optimal cooperative strategy for multi-path routing and subflow allocation in real time according to the dynamically changing network loads.It improves the transmission success rate.