锚节点位置不确定情况下的鲁棒室内定位方法研究

发布时间：2018-01-11 18:02

本文关键词：锚节点位置不确定情况下的鲁棒室内定位方法研究　出处：《哈尔滨工业大学》2015年硕士论文　论文类型：学位论文

【摘要】：室内定位的需求在未来将会不断增加,而对于定位精度的提高一直是研究的重要目标。影响定位精度的因素有很多,不管是考虑环境因素的非视距传播、多径传播,还是由噪声引起的测量距离误差,都会导致定位出现误差。针对室内定位,在考虑锚节点位置的不确定所导致的定位误差方面的问题,同样需要得到足够的重视。本文在常见的定位模型上,考虑了锚节点位置不确定问题,重新建立了本文所用到的定位模型,并对其数学建模,以便用于文中的仿真和分析。通常情况下,我们常常在仿真和应用中认为得到的锚节点位置就是其真实位置,而实际情况并非如此,在锚节点位置不确定的模型中,我们需要先获得锚节点的位置坐标,然而通过众多方式所获得的锚节点位置,会因为各种原因,使得所谓已知的锚节点位置相对于真实位置是带有误差的。借鉴测量距离的误差对定位产生影响的分析方式,先从测量距离的误差服从高斯分布的分析入手,逐渐向实际的测量结果靠拢;以及众多文献对存在定位误差作了详细的研究和分析。我们便借鉴了其中一些能够适用于分析本文所需要解决的问题的方法,可以分为两大部分:非贝叶斯估计方法和贝叶斯估计方法。采用这两种方法是因为,非贝叶斯方法中有很多算法可以用于分析锚节点位置不确定所带来的定位误差,算法常见,而且简单易行,假定的误差分布也是常见的高斯形式。如果需要考虑实际情况所得到的锚节点误差,我们便可以从贝叶斯的方法入手,这种方法更加具有一般性,而且以便于后续的研究。本文一方面通过非贝叶斯估计方法,考虑了四种线性的算法,仿真和分析了锚节点位置不确定对定位精度的影响,考虑了锚节点误差大小以及定位区域大小,得出了不同算法之间在不同的条件下定位性能的差异性,直观的反映了锚节点位置不确定所带来的定位误差大小和利用不同算法的特性来指导实际情况的应用。解决了锚节点误差导致的奇异解问题,提高了定位精度和定位性能,保证了定位算法的鲁棒性。另一方面通过贝叶斯估计方法,引入置信度传递的算法,由锚节点的先验知识求得定位的边缘后验概率分布,从参数形式转向非参数形式,用基于粒子的信息传递和分布的高斯近似,从而能够解决锚节点误差服从实际分布情况的问题,相对于非贝叶斯方法更加通用,也可应用置信度传递于锚节点自我迭代的精度提升和后续进行协作定位方面的研究。
[Abstract]:The demand for indoor positioning will continue to increase in the future, and the improvement of positioning accuracy has been an important goal of the study. There are many factors that affect the positioning accuracy, whether it is non-line-of-sight propagation considering environmental factors. Multipath propagation, or measurement distance error caused by noise, will lead to positioning errors. For indoor positioning, the positioning error caused by the uncertainty of anchor node location is considered. In this paper, considering the uncertainty of anchor node location, the location model used in this paper is re-established, and its mathematical model is established. In order to be used for simulation and analysis in this paper. In general, we often think that the anchor node position is the real position in simulation and application, but the actual situation is not. In the model of uncertain anchor node position, we need to obtain the anchor node position coordinates first. However, the anchor node position obtained by many ways will be due to a variety of reasons. So that the so-called known anchor node position relative to the real position is with errors. For reference to the measurement distance error impact on the positioning analysis, first from the error of the measurement distance from the Gaussian distribution analysis. Gradually close to the actual measurement results; And many literatures have done a detailed study and analysis of the existence of positioning errors. We have used some of the methods that can be used to analyze the problems that need to be solved in this paper. It can be divided into two parts: the non-Bayesian estimation method and the Bayesian estimation method. There are many non-Bayesian methods can be used to analyze the location error caused by the uncertainty of anchor node location. The algorithm is common and easy to implement. The assumed error distribution is also a common Gaussian form. If we need to consider the actual situation of the anchor node error, we can start from the Bayesian method, this method is more general. On the one hand, four linear algorithms are considered through the non-Bayesian estimation method, and the influence of the location uncertainty of anchor nodes on the positioning accuracy is simulated and analyzed. Considering the error of anchor node and the size of location area, the difference of localization performance between different algorithms is obtained under different conditions. It intuitively reflects the location error caused by the uncertainty of anchor node location and uses the characteristics of different algorithms to guide the application of the actual situation and solves the singular solution problem caused by anchor node error. The location accuracy and performance are improved, and the robustness of the localization algorithm is guaranteed. On the other hand, the confidence transfer algorithm is introduced through Bayesian estimation. The edge posteriori probability distribution is obtained from the priori knowledge of anchor nodes, and the probability distribution is changed from parametric form to non-parametric form. The Gaussian approximation based on particle information transmission and distribution is used. Therefore, it can solve the problem of the actual distribution of the anchor node error, which is more general than the non-Bayesian method. It can also be used to improve the accuracy of self-iteration of anchor nodes and to carry out the research of cooperative localization.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN92

【参考文献】