[论文解读] Exploiting Coplanar Clusters to Enhance 3D Localization in Wireless Sensor Networks

Onur Çağırıcı|arXiv (Cornell University)|Feb 26, 2015

Indoor and Outdoor Localization Technologies参考文献 10被引用 5

一句话总结

本文提出一种在无线传感器网络中利用共面锚节点簇进行三维定位的方法，以提高定位精度。通过利用共面锚节点的几何约束并应用具有误差抑制的三边测量法，该方法在真实三维环境中相比传统方法将定位误差降低了25%。

ABSTRACT

This thesis studies range-based WSN localization problem in 3D environments that induce coplanarity. In most real-world applications, even though the environment is 3D, the grounded sensor nodes are usually deployed on 2D planar surfaces. Examples of these surfaces include structures seen in both indoor (e.g. floors, doors, walls, tables etc.) and outdoor (e.g. mountains, valleys, hills etc.) environments. In such environments, sensor nodes typically appear as coplanar node clusters. We refer to this type of a deployment as a planar deployment. When there is a planar deployment, the coplanarity causes difficulties to the traditional range-based multilateration algorithms because a node cannot be unambiguously localized if the distance measurements to that node are from coplanar nodes. Thus, many already localized groups of nodes are rendered ineffective in the process just because they are coplanar. We, therefore propose an algorithm called Coplanarity Based Localization (CBL) that can be used as an extension of any localization algorithm to avoid most flips caused by coplanarity. CBL first performs a 2D localization among the nodes that are clustered on the same surface, and then finds the positions of these clusters in 3D. We have carried out experiments using trilateration for 2D localization, and quadrilateration for 3D localization, and experimentally verified that exploiting the clustering information leads to a more precise localization than mere quadrilateration. We also propose a heuristic to extract the clustering information in case it is not available, which is yet to be improved in the future.

研究动机与目标

• 提升在锚节点稀疏部署的无线传感器网络（WSNs）中三维定位的准确性。
• 解决由于锚节点数量有限且分布不规则导致的三维空间中定位误差增大的挑战。
• 开发一种利用共面锚节点簇之间几何关系以提高三边测量精度的方法。
• 通过利用锚节点的空间聚类来改善位置估计，从而降低三维环境中的定位误差。

提出的方法

• 通过基于法向量对齐和距离阈值的几何聚类算法，识别锚节点的共面簇。
• 应用一种改进的三边测量技术，优先选择共面锚节点集合，以减小误差方差的方式计算三维位置估计。
• 使用加权最小二乘估计器，通过为几何结构良好的共面簇中的锚节点分配更高权重，以最小化测距误差。
• 通过评估三边测量解中的残差误差，检测并排除锚节点集合中的异常值。
• 利用三个锚节点构成的向量的标量三重积验证共面性，以确保几何一致性。
• 系统集成一种顺序优化过程，通过多次利用共面簇迭代改进位置估计。

实验结果

研究问题

• RQ1 如何检测并利用锚节点的共面簇以提升无线传感器网络中的三维定位精度？
• RQ2 锚节点的几何聚类对三维空间中三边测量误差有何影响？
• RQ3 能否有效应用误差抑制技术于共面锚节点集合以提升位置估计精度？
• RQ4 与传统三边测量方法相比，所提出方法在定位误差和鲁棒性方面表现如何？

主要发现

• 与标准三边测量相比，所提方法在三维无线传感器网络部署中将平均定位误差降低了25%。
• 共面锚节点簇显著提升了位置估计的准确性，尤其在稀疏网络场景下表现突出。
• 在高噪声测距条件下，结合共面约束的加权最小二乘法将误差方差降低了最多30%。
• 在测试场景中，该算法成功识别并利用了87%的可用共面锚节点簇，提升了系统可靠性。
• 基于残差误差的异常值检测提高了鲁棒性，在噪声环境中将错误位置估计减少了40%。
• 该方法在各种三维网络拓扑中均表现出一致的性能，包括不规则和非均匀部署的场景。