Sensor Data Query Processing Algorithms In Wireless Sensor Networks

In recent years, with the rapid development of wireless communication tech-niques, microelectronics and embedded computing techniques, Wireless Sensor Networks(WSNs) are being used widely in many fields, such as environment monitoring, healthcare and military defense. A WSN is composed of large numbers of sensor nodes de-ployed in the monitored region. Each sensor node has some computing ability, storageability and communication ability. Many sensor nodes collaborate in an Ad Hoc way tosense, collect and process information obtained from the objects monitored by the WSNs.They complete one or more tasks jointly. Since WSNs can sense its around environmentsand generate various kinds of sensor data continuously, they are looked as the novel dis-tributed database systems. Users can get the information of the monitored region by issuethe queries to WSNs. However, the existing database query processing techniques arenot suitable for WSNs due to the computing ability, storage ability and communicationability of sensor nodes are all limited. So, the novel query processing techniques whichare suitable for WSNs are needed urgently. Regarding the features of WSNs, the studyof query processing techniques for WSNs was focused on by this dissertation. The maincontributions of this dissertation are as follows:First, a model-fitting based believable sensor data collection query processing al-gorithm was proposed. The sensor data collection in the monitored region is one of themost fundamental, important and widely applied application in the various applicationsof WSNs. Based on the sensor data collected, users can not only know about the informa-tion of the monitored region, but also get some research results by further processing andanalysis. Due to the characteristic of WSNs being energy limited and hard to recharge,how to collect the sensor data energy efficiently and prolong the lifetime of WSNs is achallenging problem that is urgently to be resolved. In addition, the users require the col-lected sensor data to be confident in many applications. Because these collected sensordata will be the basis for the users'further research. By now, the related works mainlyfocus on how to save energy, but not consider the confidence of the collected sensor data.So, a model-fitting based believable sensor data collection query processing algorithmwas proposed in this dissertation. The algorithm proposed can not only collect the sensor data energy efficiently, but also ensure the confidence of the collected sensor data.Second, a new query which called Probing Query was proposed and three algorithmsfor processing probing queries were given. For queries in WSNs, empty sets may bereturned as query results, which could confuse users a lot and users obtain no usefulinformation about the monitored objects from the empty sets. Modifying the queries andthen executing them again and again will not only block the execution of other queries, butalso increase the energy consumption and shorten the lifetime of WSNs. For resolvingthis problem, probing query was proposed in this dissertation and three algorithms forprocessing probing queries were given. Probing queries can return users probing resultsets that consist of the sensor data with the smallest deviation from given queries. Probingresult sets can be used not only to answer the users queries approximately, but also toprovide users reference information about modifying queries. Experimental results showthat the algorithms proposed in this dissertation performance well and are energy efficient.Third, a distributed two-stage Top-k query processing algorithm was proposed. Top-k query being as an important complexity query can be applied for many applications inWSNs. By now, the existing studies of Top-k query in WSNs are all continuous queries,but no snapshot queries. To answer the users snapshot Top-k queries, a distributed two-stage Top-k query processing algorithm was proposed in this dissertation, which considersthe energy limitation of WSNs and aims to decrease the energy consumption of WSNsmaximally. The algorithm proposed can compute a Top-k query result set energy effi-ciently by propagating the filters into WSNs.Fourth, a multiple-regression model based missing data estimation algorithm wasproposed. Missing sensor data is inevitable in WSNs due to the inherent characteristic ofWSNs, and it causes many difficulties in various applications. To resolve this problem, amultiple-regression model based missing data estimation algorithm was proposed in thisdissertation, based on the temporal and spatial correlation of sensor data. The algorithmestimates the missing data with the sensor data of multiple neighbor nodes jointly. So,the algorithm may estimate the missing data accurately. The missing data estimationalgorithm can be used to process the cases that missing data exists in query result set.The research of missing data estimation algorithms is a useful supplementation for theresearch of sensor data query processing algorithms.