Research On Localization Issue In Anisotropic Wireless Sensor Networks

Location information usually serves as the fundamental part of most wireless applications which range from a short distance bluetooth communication to long distance telephone network communication, and the data without geographical position is meaningless. However, traditional localization algorithms focus on positioning accuracy and efficiency rather than the context-aware environment and sensor hardware constraints. We summarized these as"anisotropic"and our algorithmic study of localization in anisotropic sensor networks is the main contribution in the dissertation.The dissertation targets at the localization problems and difficulties in anisotropic sensor networks. It presents in-depth study on localization problems in anisotropic sensor networks based on computational geometry theory and game theory, and propose novel research approaches to decrease error ranging and improve localization accuracy and convergence speed in anisotropic WSNs. The main contributions of the dissertation are as follows:1. We first design a range-free weight-based selective anchor nodes localization algorithm (WSAN) in Chapter 3 which targets at irregularity problem of localization area. The study of traditional multi-hop based localization problem is in the precondition of a mapping function between measured distance and Euclidean distance for pair of sensor nodes. Due to the variety of application scenarios and distributed environments of sensor nodes, if the sensors are deployed in an anisotropic network, traditional assumption of isotropic sensor network may not hold which may further lead to large errors. In the proposed WSAN algorithm, we filter out the reference nodes based on shortest paths between reference nodes and unknown nodes that are seriously affected by boundary or barrier in anisotropic area. Each unknown node will chooses the suitable reference nodes to reduce the error estimation in localization process and improve localization accuracy by choosing the reference nodes that are less affected by the anisotropic factors involved in the localization. Our evaluation results show that the algorithm can improve the accuracy level by 30% comparing with traditional localization algorithms by selecting the proper reference nodes.2. In Chapter 4, we introduce a range-free convex-hull partitioning localization algorithm CHP for anisotropic networks and a pervasive localization method with a broader applicability. At the beginning of CHP, all the reference nodes are formed into different convex-hulls. Then each unknown node keeping information and choosing the convex-hull which is less affected by the anisotropic factors involved in the localization. At last, each unknown node executes a localization process based on its convex-hull's reference nodes. The CHP algorithm can effectively reduce the localization errors incurred by the boundary or barrier factors of complex area. The results from extensive simulations show that compared with traditional algorithms, the CHP algorithm significantly reduce the localization errors and error jitters.3. A game theory based anisotropic sensor networks localization algorithm GTCMS is introduced in Chapter 5. Most of the localization algorithms comprise the following two phases: initial position estimation phase and collaborative refinement phase. These algorithms use neighbor nodes and reference nodes to refine the position of nodes estimation. Nevertheless, there has not been any paper mentioning the problem of how to refine the position of nodes in a most rational way. During the position refinement process, all nodes are willing to preserve the best position, and meantime act selfishly so as to minimize the movement distance if the position estimation of nodes is accurate enough. Game theory is a suitable tool to analyze the conflicting objectives of nodes purchasing to refine the position estimation. In the position refining procedure of our GTCMS algorithm, each node first selects the neighbor nodes that help adjust its position by following the natural game course users, then uses the best response theorem we proposed to adjust its position. We prove that the GTCMS algorithm can achieve the global Nash equilibrium. To the best of our knowledge, this is the first work in the game theoretic study of localization problems. Compared with the localization methods without using the game theory model, GTCMS algorithm greatly reduces the localization error and converges much more efficiently.4. The dissertation proposes a Routing information Correction based Localization scheme (RCL) for anisotropic sensor networks. Generally speaking, traditional localization algorithms make assumptions such as regular radio transmission mode for each sensor, each sensor could achieve symmetric communication in its radio range, and do not have packet loss problem. Nevertheless, we find the radio pattern of each sensor is not stable in most of the time based on our real experiment. Furthermore, there may exist significant radio mode difference between pair of nodes even if they have the same transmit power, and cannot achieve symmetric communication. The communication radius for each sensor equipped with omnidirectional antenna fluctuates with the changes of communication directions and time points. Besides, communication quality of wireless link between pair of sensors may be poor, especially for pair of nodes which are distant enough. These shortcomings mentioned above may seriously affect the localization accuracy in application. This part focuses on irregular communication model, time variation factor, packet loss problems, and proposes the RCL localization algorithm. We solve the above problems by introducing a series of strategies in RCL. It is demonstrated that a better shortest path construction process could be guaranteed by the proposed RCL method. With RCL, we can reduce the negative effects brought by irregular communication model in a maximum extent, and effectively improve the localization accuracy.We introduce the computational geometry approach and game theory method into the anisotropic sensor network localization issue in the dissertation, which bring new insight into this field. At the end of the thesis, we point out that there needs improvements in various aspects and present the future work.