QoS-aware query processing for wireless sensor networks

A Wireless Sensor Network (WSN) is a collection of sensor nodes, spatially deployed in an ad hoc fashion, which performs distributed sensing tasks in a collaborative manner, without relying on any underlying infrastructure support. WSNs have usually been treated as passive networks, where sensors measure some physical phenomena periodically, or detect some events, and then send the data thus generated to the sink. Another kind of application is emerging, both in the context of a stand-alone WSN and of distributed WSNs, for the Internet of Things (IoT) where, instead of serving only as passive, information gathering mechanisms, sensors can actively process and respond to user queries to report the sensed data on demand. Irrespective of the query types, query processing tasks in WSNs pose several technical challenges, such as (i) selecting the appropriate sensors, within a stand- alone WSN or the appropriate WSNs among the distributed WSNs in the IoT, which satisfy the query specifications; (ii) providing the query response within the user-specified delay bound; (iii) ensuring higher query response accuracy; and (iv) determining the query execution plan dynamically to satisfy the QoS requirements on either delay or accuracy, or a combined specification of both delay and accuracy. This thesis aims to address these issues by developing a hierarchical data model for representing the sensor data of a stand-alone WSN at sensor, fusion node and sink levels. In the case of an aggregate query, the model can be used to select the appropriate sensors which satisfy the query specifications. In the case of an approximate query, the model can provide the approximate query response directly from the model. In both instances, the data model enables retrieving the query response with higher accuracy and less data transmission. To execute a query in delay sensitive applications, a delay aware spanning tree is designed, which provides a routing path for query dissemination to the target sensors as well as collecting a response from them. The tree is constructed based on the query load of the sensors so that the sensors with a higher query load can be reached quickly thus causing less delay in retrieving the query response. Moreover, to support the query execution according to the QoS requirements on delay and accuracy, the hierarchical data model is applied on the delay aware spanning tree to generate a QoS aware sensor model. The model organises sensor data hierarchically, with different levels of accuracy, on different hop distances of the tree. It creates the provision to collect the query response from any hop distance of the tree to satisfy the query-specified delay and accuracy bound. Then, the formulations are extended for the distributed WSNs in the context of the IoT. In doing so, a distributed WSNs information model is maintained in the web server which gathers the data models and the spatial properties of the WSNs from their associated QoS aware sensor models. The information model exploits the overlapping information among the participating WSNs and facilitates selecting the appropriate WSNs it is necessary to visit according to the query specifications. Moreover, based on the QoS requirements of the query, the information model determines the cost effective query execution plan, with the help of the QoS aware sensor models of the respective WSNs. Extensive analytical and simulation analyses confirm the efficacy of the proposed new strategies compared with notable work in this area