Positioning with Android: GNSS observables
January 17, 2018 - By Stuart Riley, Herbert Landau, Victor Gomez, Nataliya Mishukova, Will Lentz and Adam Clarehttp://gpsworld.com/positioning-with-android-gnss-observables/ Summary
While this article has highlighted that sub-meter and centimeter accuracy have been achieved in a laboratory environment, there are many challenges to be addressed before centimeter accuracy in a phone can be achieved with performance suitable for users in real-world environments.
Figure 24. CDF plots for the different PPE position solutions for cell antenna dataset.
The challenges include very high multipath, significant differences in the tracking performance between different devices, and high rates of cycle slips. As very few Android-based devices provide continuous phase, a pseudorange/Doppler-based engine has been modified to accept Android data.
Based on the testing with existing devices it is possible to achieve position solutions of 1–2-meter accuracy in ideal static scenarios. This is a significant improvement in accuracy for Android based devices.
Figure 25. PPE engine on a Samsung S7 with a Broadcom BCM4775 evaluation kit.
However, as performance differences were observed between different mobile devices significantly more data needs to be collected over a larger set of devices to review the repeatability of these preliminary results from existing Android devices.
The Broadcom BCM47755 development kit for a dual-frequency GNSS chipset intended for future phones has allowed us to review the potential position performance that may be achievable in a handset in a few years.
By connecting this next-generation GNSS chipset to a GNSS antenna typical of a cellular device and comparing the performance from a precision GNSS antenna, we’ve shown for the first time that it is possible to produce precision positions from a static cellular class GNSS device in ideal conditions at the centimeter level with both an RTK solution and a PPP solution.
However, due to the significantly higher measurement noise and high multipath from the cellular device’s GNSS antenna, the convergence times to reach centimeter level remain a challenge; although using dual-frequency phase data from a cellular GNSS chipset with a PPE and RTX service, the position is very rapidly sub-meter.
Future work will focus on analyzing how the performance changes when operating in the normal user environment. The effects on the measurements of user motion, body masking and de-tuning of the antenna when the device is held need to be quantified. The Nexus 9 tablet used in this article does not have integrated cellular. The Broadcom development kit connects to the phone via a cable and is also not integrated into the handset.
We will be evaluating what may happen with a more integrated unit to make sure emissions from devices with integrated cellular very close to the GNSS antenna do not result in further degradation.
As the position performance is very sensitive to the quality of the antenna from both multipath and cycle slips due to low C/NO and deep fades, we’ll also evaluate how well the performance of the PCB-based GNSS antenna, which is part of the BCM47755 evaluation kit, matches current handsets.