Iterative Localization of Wireless Sensor Networks: An Accurate and Robust Approach

In wireless sensor networks, an important research problem is to use a few anchor nodes with known locations to derive the locations of other nodes deployed in the sensor field. A category of solutions for this problem is the iterative localization, which sequentially merges the elements in a network to finally locate them. Here, a network element is different from its definition in iterative trilateration. It can be either an individual node or a group of nodes. For this approach, we identify a new problem called inflexible body merging, whose objective is to align two small network elements and generate a larger element. It is more generalized than the traditional tools of trilateration and patch stitching and can replace them as a new merging primitive. We solve this problem and make the following contributions. Our primitive can tolerate ranging noise when merging two network elements. It adopts an optimization algorithm based on rigid body dynamics and relaxing springs. Our primitive improves the robustness against flip ambiguities. It uses orthogonal regression to detect the rough collinearity of nodes in the presence of ranging noise, and then enumerate flip ambiguities accordingly. We present a condition to indicate when we can apply this primitive to align two network elements. This condition can unify previous work and thus achieve a higher percentage of localizable nodes. All the declared contributions have been validated by both theoretical analysis and simulation results.