Analysis and Pre-processing of Raw Measurements From Smartphones in Realistic Environments

The rising need for high-accuracy location-based services in mass-market applications has boosted the demand to improve GNSS positioning in low-cost receivers such as those found in smartphones [1]-[6]. Newer applications in autonomous navigation, sports, gaming, bicycle rentals, personal navigation and fitness tracking make use of low-cost or ultra-low-cost GNSS hardware. Smartphones contain extremely low-cost GNSS chips with very low-cost antennas which are, mostly, GPS single-frequency (L1) receivers. Smartphones tend to provide locations based on a combination of GPS code-only Single Point Positioning (SSP) aided by an internal inertial management unit (IMU) and cell tower ranging information. The accuracy of such a technique is at the metre or tens-of-metres level with drifts in cases of signal loss from either satellite or cell towers. The quality of the new generation of low-cost (<100 USD) GNSS chips has improved drastically over the last few years. In 2018, dual-frequency, multi-GNSS chips made their way into smartphones, with the Xiaomi MI 8 phone being the first one, with a Broadcom BCM47755 chip. Since 2016 and with the advent of Android N phones, both carrier and pseudorange measurements were made available to the public [5], [8]. The realistic usage of smartphones lies in semi-urban, urban and forested areas with the phone in hand or on the dashboard of a car when it comes to positioning-based applications. Such usage brings with it numerous challenges and limitations such as missing measurements, poor multipath suppression as well as a low and irregular carrier-to-noise-density ratio (C/N0), which need to be analysed and addressed. Hence, this research focusses on the analysis and conditioning of raw measurements from smartphones for use with Precise Point Positioning (PPP) augmentation in realistic environments. PPP processing is preferred over relative positioning techniques for smartphones since the technique functions without any base station and baseline distance limitations. This research dwells deeply into the following aspects: 1) Analysis of the carrier-to-noise-density ratio, multipath and frequency and length of data gaps in different multipath environments and their correlation with each other; and 2) Prediction of missing measurements and its impact on the availability and accuracy of the positioning solution. The analysis indicates that in semi-urban areas, the rms multipath affecting the code observations can be as high as 16 m with an average of only 55% of the total epochs having both code and carrier-phase measurements from both frequencies present. After utilizing a C/N0-based measurement weighing strategy and prediction model, results indicate achieving nearly 100% positioning availability in semi-urban scenarios and a 35% decrease in the standard deviation of the horizontal positioning error.