基于流量模式的Q-学习路由及其连接调度

为解决无线网状网中因多条路径同时传输数据而引起网络性能降低的问题,提出了一个基于流量的Q-学习路由与调度方案(QRST):针对每一个路由请求,首先采用强化学习中的Q-学习算法寻找路径;然后根据找到的路径结合信道分配完成组合调度,以启发式的方法尽可能为每个时隙使用网络资源分配路径的连接.并在不同网络资源配置和多种流量请求...

Q-Learning Routing and Link Scheduling Based on Traffic Mode

The interference and resource congestion caused by multiple concurrent flows may cause sharp perfor-mance degradation of wireless mesh networks. In order to solve the problem, a Q-learning Routing and Scheduling concerning Traffic (QRST) scheme is proposed. Firstly, the Q-learning algorithm is used to find the path for each routing request. Then the combined scheduling is completed according to the path finding and channel allocation, and the connection of paths is allocated with cyber source for every slot in a heuristic way. In order to verify the co-rrectness and effectiveness of the scheme, virtual computing is performed under different network resource configurations and multiple traffic requests. The experimental results show that, compared with COSS and AODV, wireless mesh network using the QRST scheme has better performance in terms of throughput, activated link number and transmission completion time.