Research On Localization In Wireless Sensor Networks

In recent years, with the in-depth research of networking techniques, wireless sensor network is widely used for bridging the physical world and the computer world, which shows substantial research and application importance. Localization is one of the key issues in the design and implementation of wireless sensor networks. Localization provides location service for the entire network, which is a supportive technique in wireless sensor networks.Currently, many researchers work on the localization problem, and propose lots of localization algorithms, which obtain fairely good results in their own scenario under certain reqirement. Nevertheless, most existing algorithms suffer the following two problems. First, they only focus on the research of single-node localization, and neglect the possible collaboration of the target nodes. Most current localization algorithms adopt single-device-localization algorithm to solve network localization problem, which localize each unlocalized node by single-node localization technique. However, this method neglects the possibility of the potential collaboration between the unlocalized nodes, which is over-simplified for network localization. In fact, there are many nodes that cannot be localized by single-node localization technique, but can be localized by collbarative localization, which are missed by traditional algorithms. Hence, the performance of localization needs further research. Second, the error control cannot fulfill the demond of the applications. As the distance measurements always contain errors in practice, many literatures investigate the impact of the measurement errors for the accuracy of localization result to obtain robustness. Unfortunately, these algorithms are too conservative, so that they wrongly indentify many reliable results as non-robust. Therefore, the robustness condition prevents many robust results, leading to low localization performance.This paper systematically investigates some key issues of localization in wireless sensor networks, especially on the localization performance and robustness, which are the most important requirements of localization algorithms. The research mainly focuses on the following aspects.For hig-performance localization, this paper breaks the inherent limitation of node-based design, and proposes component-based localization. Existing node-based designs adopt node as the basic unit for localization, and achieve the utmost performance in this subject. Nevertheless, there are many localizable network instances that cannot be localized by node-based algorithms. To address this issue, this paper proposes a component-based design. Component-based algorithm utilizes components, instead of nodes, as basic units for localization, which are defined as a two-dementional rigid structure. By intergrating into components, nodes can better share their distance measurements as well as anchor information, thus to be localized as a whole. This paper builds the theoretical foundation of component-based localization, and futher introduces the design of component-based algorithms. Moreover, the gain of performance by component-based design is also analyzed. By introducing a new metric, this paper proves that the component-based localization achieves the highest performance amonge existing algorithms.Besides proposing component-based localization, this paper further investigates robust component-based localization, when using noisy distance measurements. Basic component-based localization is based on the idealized model, which assumes accurate inter-node distance measurements. However, this model is over-idealized for current measurement techniques. To cooperate the component-based design with the measurement errors, this paper propose robust component-based localization algorithm to obtain robustness in localization. By introducing several patterns for robust coordinate system conversion, this paper properly solves the challenge brought by measurement errors. Robust component-based localization technique not only inherits the high performance of the component-based design, but also gurantees the robustness of the result.For robust localization, this paper further proposes the concept of strongly localizable network. Traditional robust localization mainly focuses on the error distribution of the result and the structural properties of the network, making the analyses only fit for the certain scenario and certain problem without any generality. To conque such difficulties, this paper propose the concept of strongly localizable network to investigate generic robust localization probolem. A formal definition of robustness is given by the terminology of localization theory. In addition, this paper proposes a novel algorithm to identify and localize a special kind of strongly localizable network: robust trilateration network. Through equivalent trasformantion of the ranging errors, the impact of the errors is bounded, thus to propose a concise and effective way to obtain robustness in localization. Compared with existing algorithms, this technique is more precise in robustness prediction, so that it can reduce the false positive rate. Therefore, the proposed method obtains higher localization performance than exsiting designs.Based on the research of robust trilateration network, this paper proposes an optimistic localization scheme to improve the design of robust localization mechanism. Existing robust localization algorithms often adopt pessimisitic mechanism, which obtains robustness by dropping the localization results failing in the robust test. However, pessimistic mechanism cannot gurantee to identify all robust results by the robust test, so that this mechanism may greatly degrades the localization performance. To address this problem, this paper proposes the concept of optimistic localization. In contrast to traditional design, optimistic algorithm first assumes that the localization result is robust, and then checks the possible errors in the result set by the consistency of the position assignments. After indentifing errors in the result set, optimistic algorithm can also correct the wrongly localized nodes. In a word, optimistic algorithm not only preserves the robust results, but also changes the incorrect results into robust ones. As a result, the performance of the optimistic algorithm is much higher than the existing robust localization schemes. Though a problistic model, this paper also analyzes the lower bound of the performance gain as well as the cost of optimisitic design. The result shows that optimistic algorithm has rigorously higher performance than the traditional ones.To summarize, this paper investigates two essential issues of localization: performance and robustness, which are key requirements in wireless sensor networks. The formal analyses and extensive simulations all show that the proposed algorithm can properly achieve the design goals.