Integration of signal prediction service in Service Oriented Architecture

Abstract

Smart transducers based on the IEEE 1451 standard provide platform for development of distributed measurement and control systems. Integration of IEEE 1451 services based on Service Oriented Architecture (SOA) enables simple implementation, interoperability, scalability of distributed system components, accessible in the form of network services. Addressing the transport latency and packet loss in distributed systems, we propose the concept of algorithms' and processes' implementation in the form of dislocated network services. Introducing the signal data paths, time-stamped data and time synchronization mechanism, we provide the support for real-time system characteristics. Each component on the data path is able to analyze the packets of time-stamped information, extract an adequate sample set, and apply this set to the input of the implemented real-time algorithm or actuator. Proper management of the time-stamped information along data path enables real-time actions despite the network propagation delay or other systems' latencies. In this paper, the presented architecture is used for integration of the signal prediction service in smart transducers network based on SOA. Purpose of the service is to predict the values of arbitrary signals in order to compensate network delay and packet loss. Two realizations of the signal prediction service, based on neural networks are given: feed-forward and feedback predictive network. The system containing embedded IEEE 1451 smart transducer module for relative humidity measurement and signal prediction service on the general purpose platform is used for the experimental verification of the proposed concept.