Localization In Wireless Sensor Networks

Localization is an important technique in wireless sensor network, as location is not only the indispensable information in the data collection system, but also assists the design of the network protocol. This paper first introduces the current emerging localization algorithms in wireless sensor networks, then for the localization in static and large scale wireless sensor networks, we study deeply on the multidimensional scaling algorithms which is suitable for this condition, for the device free tracking problem, we propose an algorithm based on particle filter.The main work in this paper is as follows:Primarily, we specify the advantages and disadvantages of the algorithms based on multidimensional scaling. Then we propose an improved distributed multidimensional scaling algorithm, which increases the localization accuracy and coverage speed. Firstly, in order to overcome the inefficiency of the initialization phase of the original algorithm, we improve the method of initializing the nodes locations, which accelerates the coverage speed. Secondly, we propose a strategy of neighbor choosing to solve the flip problem brought by the insufficiency of anchor nodes, which helps to lower the probability of the node flipping and prohibit the propagation of the large error due to flipping. Finally, we make a detection of node flipping and further adjust the locations of the nodes. All of these strategies could help the decrease of localization error.Then, experiments for multidimensional scaling algorithm have been implemented in this paper. We first measure the distances between nodes based on NANOPAN5375in the real world, which gives us the statistical parameters of the distance measurement noise. Then these parameters are used in the simulation. The simulation results show that our new algorithm has a lower localization error when comparing with the classical multidimensional scaling algorithm and the distributed weighted multidimensional scaling algorithm in uniform distribution, C-shape distribution and random distribution. In the condition of random distribution, the localization error of our algorithm increases as the increase of the noise variance of distance measurement, decreases as the increase of the node density and anchor density. Except the condition of low anchor density, our algorithm outperforms other two algorithms in localization error.In addition, for the device free tracking problem, we gives a tracking algorithm based on particle filter and then analyses the effects of parameters on the localization error. Results show that this algorithm could track the target effectively. And the localization error increases as the weight response distance threshold increases, first decreases then increases as the weight response fading threshold and the maximum speed threshold increases.To summarize, this paper studies multidimensional scaling algorithm in depth and proposes some simple and effectively improvements for distributed weighted multidimensional scaling algorithm. Moreover, we solve the device free tracking problem effectively based on particle filter.