Clustering and Representation of Time-Varying Industrial Wireless Channel Measurements

The wireless devices in cyber-physical systems (CPS) play a primary role in transporting the information flows within such systems. Deploying wireless systems in industry has many advantages due to lower cost, ease of scale, and flexibility due to the absence of cabling. However, industrial wireless deployments in various industrial environments require having the proper models for industrial wireless channels. In this work, we propose and assess an algorithm for characterizing measured channel impulse response (CIR) of time-varying wireless industrial channels. The proposed algorithm performs data processing, clustering, and averaging for measured CIRs. We have deployed a dynamic time warping (DTW) distance metric to measure the similarity among CIRs. Then, an affinity propagation (AP) machine learning clustering algorithm is deployed for CIR grouping. Finally, we obtain the average CIR of various data clusters as a representation for the cluster. The algorithm is then assessed over industrial wireless channel measurements in various types of industrial environments. The goal of this work is to have a better industrial wireless channels representation that results in a better recognition to the nature of industrial wireless communications and allows for building more effective wireless devices and systems.