Characterisation and Modelling of the Indoor Pseudorange Error using Low Cost Receivers

Kurzfassung

The Global Positioning System (GPS) has made outdoor localization broadly available and is commonly employed for automotive navigation by a very large number of users. Low cost devices are currently capable of tracking the GPS C/A signal down to a power of -159 dBm without the need of assistance. With these high sensitivity devices, tracking Global Navigation Satellite Systems (GNSS) signals indoors becomes possible. However, the building material used and the architecture greatly influences the signal propagation. For instance, European houses are more affected by attenuation as they are now mostly built of concrete compared to many American buildings which are often built of wood and bricks. Typical wave propagation mechanism into most European buildings is through windows as ordinary glass attenuates electromagnetic waves less than thick concrete walls. Bearing this in mind, diffraction at window frames, as well as outdoor and indoor reflections play an important role in indoor GNSS positioning through their effects on the propagation channel. In this contribution we will try to characterize and build a model for the GPS ranging error that is caused by multipath propagation and diffraction/refraction. Since a room and its major fixtures such as furniture and windows, as well as the surrounding building(s), are largely static, a temporal correlation of the pseudorange error measured at a fixed point is to be expected. This is taking into consideration that the satellite’s motion is too slow to cause sudden channel changes due to changes in the incoming wave angles. Additionally, a certain spatial dependence could be expected as reflecting objects such as walls will remain in their position. However, due to the interaction of different reflected waves a small change in receiver position might degrade the spatial error covariance. Our main area of interest is indoor/outdoor pedestrian navigation using low cost sensors available in the mass market. Accordingly, in order to investigate the pre-mentioned effects and build a pseudoranges error model for indoor environments, we performed a measurements campaign with low cost GPS receivers that are using the commercial Sirf III chipset. Five receivers placed with a certain separation were used simultaneously in order to measure spatial correlation of the pseudorange error. Measurements took place at several positions located in a large entrance hall of a three-story building. At each measurement point the separation was varied from 20cm to 5m. As ionospheric and tropospheric errors were corrected using an additional GPS receiver placed on the roof of a nearby building, the error sequences are mainly resulting from wave propagation effects and thermal noise. The later, of course, will be essentially uncorrelated over time and space. In the paper, statistical approaches will be used to characterise and model the pseudorange error. We believe that our results are of special interest for designing positioning estimators such as Sequential Bayesian estimators. In fact, if such coloured error processes are included in the dynamic system model of these estimators, their overall positioning performance will be improved. For results in the area of multi-sensors fusion in order to combine GNSS sensors with other sources of information like Inertial-Navigation sensors, mobile radio positioning, or map matching techniques.