Tensor-based software-defined internet of things

IoT exhibits characteristics of the presence of diverse physical sensing and actuating devices, complex wireless communication and networking technologies, as well as large-scale heterogeneous data generated in the physical and cyber worlds. The exponentially increasing volume of data places an unprecedent burden on the network infrastructure of IoT systems, where there are two key challenges: how to represent the heterogeneous IoT data as a concise and unified model, and extract the essential core data that are smaller for transmission but consist of the most valuable information; and how to globally and flexibly control the network devices, and dynamically reallocate the bandwidth to improve the communication link utilization ratio. To address the mentioned challenges, this article first transforms structured, semi-structured, and unstructured IoT data to a unified tensor model, and employs the HO-SVD approach for extraction of the high-quality core data. Then this article applies SDN technology to IoT for device management, and develops a transition tensor model for routing path recommendation. Finally, a smart home case study is investigated, which reveals that the proposed tensor-based software defined model is feasible and promising. It is strongly suggested that further study on combination of IoT with SDN technology and tensor algebra should be performed.