On-Line RSS Calibration Method Based on Partial Errors-in-Variables Model

Abstract

The received signal strength-based localization with the propagation model is a simple and inexpensive method to estimate the location in an indoor environment. However, this approach gives inaccurate results due to environmental and situation-dependent noisy signal. Therefore, it is important to calibrate parameters of the propagation model in an on-line manner. In wireless sensor networks (WSNs), there exist not only uncertainties such as inherent measurement noise and inaccurate location of unknown nodes, but also certainties such as accurate location information of anchor nodes. By observing that both such certainty and uncertainty of the WSN can be appropriately formulated under the partial errors-in-variables (EIV) model, the proposed calibration method is derived under the partial EIV model to estimate the model parameters with a high accuracy. The solution can be obtained in a computationally efficient way by applying the weighted total least square method. Various simulations and real experiment are conducted to illustrate the performance of the proposed partial EIV-based calibration method, and we conclude that its performance is close to the results of the true parameters under various conditions.