Ji Ruiwen, Cao Yuanlong, Fan Xiaotian, Jiang Yirui, Lei Gang, Ma Yong
School of Software, Jiangxi Normal University, Nanchang 330022, China.
Department of Computer Science and Engineering, University of Bologna, 40126 Bologna BO, Italy.
Sensors (Basel). 2020 Nov 18;20(22):6573. doi: 10.3390/s20226573.
With the development of wireless networking technology, current Internet-of-Things (IoT) devices are equipped with multiple network access interfaces. Multipath TCP (MPTCP) technology can improve the throughput of data transmission. However, traditional MPTCP path management may cause problems such as data confusion and even buffer blockage, which severely reduces transmission performance. This research introduces machine learning algorithms into MPTCP path management, and proposes an automatic learning selection path mechanism based on MPTCP (ALPS-MPTCP), which can adaptively select some high-quality paths and transmit data at the same time. This paper designs a simulation experiment that compares the performance of four machine learning algorithms in judging path quality. The experimental results show that, considering the running time and accuracy, the random forest algorithm has the best performance in judging path quality.
随着无线网络技术的发展，当前的物联网（IoT）设备配备了多个网络访问接口。多路径TCP（MPTCP）技术可以提高数据传输的吞吐量。然而，传统的MPTCP路径管理可能会导致数据混乱甚至缓冲区阻塞等问题，这严重降低了传输性能。本研究将机器学习算法引入MPTCP路径管理中，提出了一种基于MPTCP的自动学习选择路径机制（ALPS-MPTCP），该机制可以自适应地选择一些高质量路径并同时传输数据。本文设计了一个模拟实验，比较了四种机器学习算法在判断路径质量方面的性能。实验结果表明，考虑到运行时间和准确性，随机森林算法在判断路径质量方面具有最佳性能。
