Study On Key Techniques Of Real-time Data Management In Wireless Sensor Networks

Modern wireless sensor networks (WSNs) has been improved a lot due to the development of sensor technique, micro electromechanical system, modern network and wireless communication. People can obtain physical information by dispose wireless sensor networks in certain places that human can't get close to. Therefore, it will play an important role in many areas such as military affairs, environment monitoring, industry control, city management etc. In all these applications, sensor nodes collaborate with each other to perceive, collect and process the data from the objects in the covered network area, then send the results to users for further analysis by multi-hop wireless communication. Therefore, data management is the key technique in WSN since it's a data-centric system.Real time assurance is one of the most important applications of WSN. For instance, the system needs to provide rapid response time in applications such as disaster relief, military attack and defense. This dissertation studies the key problem of real time data management in WSN. Our main contributions include:(1) A framework for real time data management in WSNs is built. We study the main real time feature of data, communication and query respectively and discuss some techniques that guarantee the real time data management. And the real time requirement will be fulfilled in lots of aspects such as data model, query language, data storage and query processing etc.(2) A data storage approach based on Multi-level mapping Index (MIS) is proposed to support both periodic query and Ad hoc query in large scale wireless sensor network. Considering the constraint of response time for periodic query and network topology, this method divides the whole network into three layers. Each layer uses local storage, data-center storage and external storage respectively. By constructing index for each layer, the low-level data can be mapped into high-level index by multi-level index. The middle level can access the low-level data in time and restrain the transmission of data with little changing. In this way, delay constraints and energy will be satisfied according to different queries. This method will save huge energy compared with local storage, exterior storage and data-centric storage..(3) Based on MIS approach, a buffer-based real-time query processing method is suggested to solve the problem that lots of queries will miss the deadline because of long distance transmission in large scale network. A query buffer is built to store data and the query mapping window is adjusted by the deadline in order to fulfill the real time requirement. The buffer-based method abstract the query initiation, query processing and result output into the queue model. The buffer size is selected according to the input speed and service ability and the extent of each network layer can also be regulated dynamically. This approach is much more efficient in energy saving and deadline satisfaction compared with method that partition the network averagely.(4) A slack factor based real-time query processing algorithm is produced to support real time and energy-saved query processing. The concept of slack factor is proposed to represent the constraint extent, meanwhile the concept of data transfer mode is proposed according to the query selectivity and the router strategy. The efficient real time query plan is established by these factors. Firstly, the location and order of query will be decided by selectivity and the hop threshold will be calculated by the slack factors. Then all the nodes will be divided into different sets by comparing the relation between the hop count and the threshold. After that each set can adopt the same data deliverty mode. In this way, the query accuracy will be improved before deadline and the energy will be saved greatly compared with the approach that only makes use of inner-network calculation or centralized calculation.(5) An efficient event detection technique based on temporal-spatial correlation model is proposed to assure the precision of detection result. Meanwhile, the constraint of response time can be satisfied by changing the execution time dynamically. The temporal-spatial correlation mode includes two processes:temporal correlation estimation and spatial correlation estimation. The temporal correlation is used to cancel the infrequent temporal mistakes and the spatial correlation is used to cancel the permanent mistakes generated by broken nodes. Taking the feature that WSN has limited energy into account; the group will collaborated by alternate work to enhance the validation time of network.(6) We design and implement a wireless sensor network data management stimulation system (WSNDM) based on the real-time data management framework and the exiting research work. Compared with the prototype system, WSNDM system integrates lots of necessary operations such as memory, index, query and schedule in WSN management. Besides implementing and validating all the methods and theories we proposed, we also accomplish many current technique for comparison.In summary, this dissertation dedicates to study fundamental problems related to data model, query processing, real time event detection in WSNs data management and proposes corresponding technical solutions. Theoretical analysis and stimulation experiments show that such methods could efficiently guarantee real time query processing and event detection. We hope that these approaches and techniques could make some referential values to develop high-performance WSN real time data management systems.