An Efficient Radio Map Construction Method For WiFi Positioning

Recently, as the rapidly increasing requirements of location based services, suchas positioning, tracking, and navigation, the positioning issue has been extensivelystudied. The fingerprint-based method is considered as a potential method withacceptable positioning accuracy in WiFi positioning area, which previously observedusing the signal strength as a scene signal characteristics to infer the position of theobserver and overcome the shortcomings of the inaccuracy of signal propagationmodel.Constructing a fingerprint radio map is essential for positioning. Most of theexisting indoor positioning methods constructed the radio map using the statisticalsampled data directly. Generally, high positioning accuracy can be obtained by adense radio map. This lead to significant effort to collect enough measurements ofreceived signal strength. Interpolation methods try to decline the number of referencepoints to construct a dense radio map, but effect is not ideal. Therefore, a key issue forthe application of fingerprint method is how to construct a radio map efficiently, andget a accurately positioning result.To achieve dense fingerprint measurements and reduce the calibration efforts,this text proposed an efficient radio map construct method based on low rank matrixcomplement theory. As an extension of compressive sensing theory, the low-rankapproximation method has been proved having high efficiency and good performancefor data recovery. The RSS data has high spatial correlation and the signal distributionof every AP can be regard as a low-rank sparse matrix. Additionally, to deal with theinterference of noises and get practical results, we revised the basic low-rankcomplement model by combining with the signal spatial consistency, namelysmoothing low-rank matrix completion model. Moreover, the singular valuedecomposition method was used to solve the proposed model.To verify the validity of the proposed method, both simulation and field test datawas used to construct radio map and positioning test. The results showed that thenumber of measurements can be significantly reduced and the better positioningaccuracy can be obtained.