Wireless Sensor Networks (WSNs), an intelligent system which is integrated withinformation collection, information transmission and information processing, have abright future in various applications and research areas. Due to the application-specificcharacters of WSNs, Quality of Service (QoS) support is indispensable to be considered.Taking the inbreak detection as an example, in order to take action in time, the localiza-tion, classification and tracking of a moving object usually needs a reliable and real-timeinformation transmission between sensor nodes and command nodes. Hence, how toguarantee the QoS requirements of the various of applications and at the same time effec-tively utilize the whole network resources deserve researching. Unfortunately, traditionalQoS guarantee techniques are not suitable for WSNs due to their unique characteristicssuch as extremely resource limitations, large scale random deployment and novel data-centric communication protocols. Furthermore, WSNs have their own particular QoSrequirements which are different from wireless networks. Therefore, in this dissertation,wefocuseontheQoSguaranteetechniquesspeciallyforWSNs. WefirstanalyzetheQoSrequirements and challenges from a wide variety of applications in WSNs, and then givean review of the current research results on QoS guarantee techniques. Next, we go deepinto the QoS guarantee problem in some aspects of WSNs, including application layer,data management layer, coverage and connectivity maintenance layer, and routing layer.The main contributions of the dissertation are as follows:First, an energy-efficient and QoS-guaranteed task allocation algorithm for homo-geneouswirelesssensornetworksisproposedinthisthesis. Withthesevereresourcelim-itations, individual sensor can not afford the computation-intensive task, multiple sensorscan accomplish a common task by collaboration. Task allocation has an important impacton the performance of real-time applications which require energy-efficient and real-timeat the same time. Due to the unique characteristics of WSNs, the existing methods fortask allocation in traditional distributed system can not be migrated to WSNs directly.Therefore, we propose a novel genetic-based nested optimal technique to guarantee theabove QoS requirements, optimal solution can be achieved by incorporating GA-basedtask mapping, GA-based routing, communication scheduling and dynamic voltage scal-ing (DVS).Second, an energy-efficient and QoS-guaranteed task allocation algorithm for het- erogenous wireless sensor networks is proposed in this thesis. In order to find an opti-mal task allocation schema that minimize the overall energy consumption while meetinguser's deadline, we exploit the divide-and-conquer technique to shrink the large searchspace. We first group all the tasks of an application into several task partitions, nextwe optimally solve the scheduling problem in branches with several sequential tasks bymodeling the branches as a Markov Decision Process. Sensor's failure can be handledby rescheduling part of the task graph. Experiment results show our proposed methodsignificantly improve the performance of sensor network in terms of energy saving andQoS-guarantee compared with two other heuristic algorithms.Third, a node sleeping scheduling algorithm combined with QoS guarantee is pro-posed in this thesis. In resource scarce WSNs, one challenge problem is to obtain longsystem lifetime without scarifying quality of service such as sensing coverage and dataintegrity. Scheduling sensors to work alternatively can prolong lifetime efficiently. Inthis thesis, we propose a data-driven sleeping scheduling mechanism which can extendlifetime by identifying redundant nodesbasedon time-spatialcorrelationsamongsensingdata. The centralized, semi-distributed and full distributed node scheduling algorithmsare presented for both small scale and large scale wireless sensor networks. Experimentresults show that our methods can prolong lifetime substantially while maintaining dataintegrity.Fourth, we take consideration of the QoS problem in query processing. An energyefficient continuous k nearest neighbors (CKNN) query processing algorithm is proposedin this thesis. Object tracking is one of the important applications of WSNs and CKNNis an essential class of spatial query in such kind of applications. In order to reduceenergy consumptions and always report the up-to-date results, we propose a two-phasessearch algorithm: initial search gets the kNN in static space and incremental maintaingets the update results during query period. The grid storage method avoids the largecommunication costs of a central storage method. A circle search method and threshold-based incremental maintain algorithm are proposed to reduce the number of messagetransmissions. Experiment results show the effectiveness and efficiency of our approachin terms of energy and latency.Finally, a QoS routing protocol which is called REER for real-time communica-tions in WSNs is proposed in this thesis. REER achieves the desired tradeoff amongcommunication delay, energy consumption, and network capacity by adapting the trans-mission power of relay nodes based on required communication delays. When deadlinesare tight, REER trades capacity and energy for shorter communication delay by increas- ing the transmission power. Conversely, when the deadlines are loose, REER lowersthe transmission power to increase throughput and reduce energy consumption. REERconsists of dynamic velocity assignment policy, forward policy, neighbor manager andqueue manager. Experiments results show that REER significantly reduces the numberof deadlines missed and energy consumption and makes a good trade off between thethroughout of non-real-time traffic and the delay of real-time traffic. |