Research On Wireless Sensor Networks Oriented Publish Subscribe System

Wireless Sensor Networks (WSN) consists of sensor technology, communication technology and intelligent service applications, which are used for extracting, transmitting and processing information of the object to be measured. Information dissemination plays as an important role in wireless sensor networks, while Publish/Subscribe (Pub/Sub) system, a middleware with decentralized control, dynamic, autonomous and loosely coupled characteristics, can provide a good platform of communication and interaction for all the distributed participants to publish events and subscribe subscriptions. The thesis does research on WSN oriented Pub/Sub technology, and develops a prototype of WSN oriented Pub/Sub system, short for JTang-WSNPS (JTang Wireless Sensor Networks oriented Publish Subscribe system). Most of WSN oriented applications are based on "push" approach, such as environmental monitoring system, intelligent agriculture, and so on. Pub/Sub model, a classical system based on the "push" approach, initiatively processes information and sends data to the clients. In traditional Pub/Sub systems, events are published by distributed senders, while in WSN applications, data with the format of data stream, is created continuously, which brings new challenges to the traditional Pub/Sub systems.Thesis does research on information processing and matching in WSN oriented Pub/Sub system. First it will get multiple data streams from wireless sensor networks, then access to multiple data streams through the proposed multi-stream processing and integrating algorithms, after that, distribute the data processing tasks to the processing servers, on which the user's subscriptions will be matched with the data.The main contributions and innovations are:Thesis proposes a strategy for constraint consolidation used to reduce the number of data transmission in WSN applications, thereby reduces energy consumption. Server collects the user's subscription request, computes constraint set for each attribute of each node, send constraint changed to the appropriate node, so the node only needs to send data within the constraints.An improved sliding window method to integrate multiple data stream processing strategy, balance the time delay and data loss rate, is proposed in the thesis. Thesis uses this improved sliding window, with both the fixed period of time sliding window's characteristics and fixed length of data sliding window's characteristics, samples the data outside of the fixed period sliding window, which maintains the data's continuity.A matching algorithm in data stream based publish subscribe system is proposed in this thesis. The algorithm merges all subscriptions'overlapping properties into the total subscription figure, orders the value of subscriptions'properties to lists, while the events as task stream are distributed by the central server, generated by the sensors.