Abstract
Recently, the proliferation of indoor location aware applications, such as healthcare, industrial guidance, and shopping mall management, has attracted enormous research interests in indoor positioning techniques for wireless sensor networks. Conventional positioning approaches include geometrical triangulation and location fingerprinting; for indoor applications, fingerprinting performs better than geometrical triangulation, but the former has to build up a huge dataset and involves much more computational complexity. In this thesis, we propose a positioning method based on propagation modeling for indoor wireless sensor networks. We use nine sensor nodes uniformly deployed in a grid (eight on the edge and one at the center) to estimate the location of a target. Each sensor measures the power of the signal transmitted from the target, and estimates the distance based on a propagation model. We can obtain six estimated (x, y) coordinates of the target by measuring the distance between the target and each of the eight sensors on the edge of the grid. After this, we remove two outliers of the six x-axis values and the six y-axis values, respectively, and then average the remaining four components in each axis to form a location estimate of the target. Based on the ninth sensor at the center of the grid, we generate another location estimate of the target. By properly combining these two estimates, we can have a more accurate positioning result. As compared to the location fingerprinting method, the proposed positioning scheme can provide better performance in location determination and tracking with less computational complexity in general. Computer simulation results are given to demonstrate the effectiveness of the proposed approach.