一种基于BP神经网络的MQTT自适应心跳机制分析

Design of an Adaptive Heartbeat Mechanism for MQTT Based on BP Neural Network

阐述通过改进霍尔特平滑网络状态检测算法,对网络时延的历史数据进行采样。使用改进BP神经网络算法对历史数据进行训练,对未来网络时延的时间点进行预测。在预测时延时间点上,使用改进MQTT心跳算法,避免连接失效,有效减少了因使用二分查找法而对MQTT协议增加额外开销。

This paper describes that the historical data of network latency is sampled by improving the Holt smoothing network state detection algorithm,the historical data is trained using the improved BP neural network algorithm to predict the future network latency time points;and at the predicted latency time points,the improved MQTT heartbeat algorithm is used to avoid connection failures.It effectively reduces the additional overhead on the MQTT protocol caused by the use of binary search method.