A Tensor-based Data Representation and Processing Framework in Cyber-Physical-Social Systems

点击翻译按钮获取中文摘要

The booming growth and rapid development in embedded systems, wireless communications, sensing techniques and emerging support for cloud computing and social networks have enabled researchers and practitioners to create a wide variety of Cyber-Physical-Social (CPS) Systems that reason intelligently, act autonomously, and respond to the users' needs in a context and situation-aware manner. The CPS systems are the integration of computation, communication and control with the physical world, human knowledge and sociocultural elements. This is a novel emerging computing paradigm and has attracted wide interests from both industry and academia in recent years.Generally, CPS systems collect massive data (Volume) from the physical world by various physical perception devices (Variety) in structured /semistructured/unstructured format and respond to the users' requirements immediately (Velocity) and provide the proactive services (Veracity) for them in physical space or social space. These collected big data are normally high dimensional, redundant and noisy, and beyond the processing capacity of computer systems. With the rapid development of CPS systems, a novel data representation and processing framework should be urgently devised to cope with the increasing large scale and high dimensional data.This proposal describes a data representation and processing framework in CPS systems based on tensors, a type of high dimensional matrix widely used in many applications. First of all, we project the CPS data onto a 6-order tensor which includes time, spatial location, cyber resources, as well as users, and establish a unified data representation model with the integration of cyber, physical and social spaces in CPS systems. Then, the incremental distributed Higher-Order Singular Value Decomposition (HOSVD) scheme based on the effecient Lanczos method is proposed to perform the dimensionality reduction and quick preprocessing on the offline/online streaming tensor-based CPSS big data. The scheme will be suitable for any devices and any processor/machines easily participate in big data processing at any time. New scheduling algorithms for various machines/processors and devices' task mapping will be investigated, as well as the corresponding case studies in some applications such as smart home and traffics to validate the feasibility and flexibility of the proposed framework.