GNSS-based calibration of the infrastructure of the repealite indoor positioning system
Résumé
In order to achieve indoor positioning, and hence a real continuity of the positioning service in all environments, we designed a new pseudolite-based system using so-called repealites. A single GNSS-like signal is transmitted from all the repealites (typically 4 for 3D positioning): this approach simplifies both the synchronization process (between transmitters) and the indoor interferences between pseudolites. In addition, in order to avoid intentional artificial multipath, the signals from two repealites are shifted in time by a few chips. This approach has already been described in previous conferences. The optical fibers, used in order to both reducing the size and the power losses of the infrastructure, allow us also to implement the time shifts and the links from the signal generator and the transmitters. Among other data required at the indoor receiver's end (locations of the transmitters for instance), values of the time shifts are required in order to help the receiver to "locate" the correlation peaks of the various transmitters. In this paper, we want to estimate the error of the GNSS receiver with respect to the reference calibration of the infrastructure carried out through the use of an optical method that has been improved during the last year. This reference calibration is fully described in the paper and leads to better than a decimeter accurate estimation of the real delays between the transmitters. Note that these delays lie typically between a few micro-seconds to a few tens of micro-seconds. The technique of estimating the biases of the receiver with respect to this calibration is based on the measurement of the delays when the outputs of the transmitters are all mixed directly to the receiver input (using microwave devices such as combiners and cables). The remaining error is in the decimeter range for absolute values and in the cm range for relative ones. This latter residue is the one of interest in our case where real measurements are differential ones indeed. High accurate positioning is thus not jeopardized