面向物联网应用的压缩感知异常数据聚合机制研究

Research on Compressed Sensing Abnormal Data Aggregation Mechanism for Internet of Things Applications

在工业应用中,作为异常检测的基础设施,物联网(Internet of Things,IoT)需要连续地收集感知数据以对工业环境进行实时监控。物联网设备通常在计算能力、存储空间、电池能量等方面都是受限的。通常情况下,工业应用的环境都处于健康的状态,异常偶尔才会发生。正常情况下,感知数据的精确值可能是没必要的,感知数据所属的区间足以支持异常检测。只有当检测到异常发生后,感知数据的精确值才有必要上传到云服务器,用于确定事件边界等异常的详细信息。针对该问题,提出了一个节能的感知数据聚集机制来支持异常检测。通过利用压缩感知算法压缩感知数据的类别数据,并在云服务器利用数据预测模型预测未来时隙的感知数据。只有当预测值的类别不同于边缘端物联网设备上传的类别数据时,物联网设备的感知数据才有必要上传到云服务器以实现云边数据的同步。为了评估该算法的有效性,做了大量的模拟仿真实验。评估结果显示,所提方法在网络流量和能耗方面相比最新的技术性能更好。

In industrial applications,as the infrastructure of anomaly detection,the Internet of Things needs to continuously collect sensing data to monitor the industrial environment in real time.Internet of Things devices are usually limited in computing power,storage space,battery energy and so on.Generally,the environment of industrial applications is in a healthy state,and anomalies occur occasionally.Under normal circumstances,the accurate value of the sensing data may not be necessary,and the interval of the sensing data is sufficient to support anomaly detection.Only when an anomaly is detected,it is necessary to upload the accurate value of the sensing data to the cloud server to determine the details of the anomaly such as the event boundary.To solve this problem,an energy-saving sensing data aggregation mechanism is proposed to support anomaly detection.The category data of the sensing data is compressed by using the compressed sensing algorithm,and the sensing data of the future time slot is predicted by using the data prediction model in the cloud.Only when the category of the predicted value is different from the category data uploaded by the Internet of Things devices in the edge,it is necessary to upload the sensing data of the Internet of Things devices to the cloud server to realize the synchronization of data in the cloud edge.In order to evaluate the effectiveness of the algorithm,a large number of simulation experiments have been done.The evaluation results show that the proposed method has better performance than that of the latest technology in terms of network traffic and energy consumption.