Page 113 - 卫星导航2021年第1-2合期

Shinghal and Bisnath Satell Navig (2021) 2:10 Page 16 of 17

Table 9 Comparison of RMS and std of positioning error for PPP, Conclusions and future work
RTK and internal positioning solution for Xiaomi MI 8, DOY 85, Tis research attempts to address gaps in GNSS research
2019 related to smartphones by focusing on analyzing and
Processing scenario Horizontal Solution availability (%) addressing challenges posed by the non-continuous and
Xiaomi MI 8 positioning error (m) poor quality of raw measurements collected in realis-
RMS STD tic environments. Te York-PPP engine has been cus-
tomized for smartphone GNSS measurements through
PPP 6.84 5.20 2443/2443 = 100 improved C/N -based stochastic modeling and measure-
0
RTK 4.55 2.80 2253/2443 = 92 ment prediction. Tested in realistic static and scenarios,
Internal solution 42.33 16.21 2443/2443 = 100 positioning accuracy is compared with results from SPP
processing, internal smartphone positioning solutions
and RTK processing. In static testing, the PPP RMS in the
solved by prediction in the PPP processing. Figure 18b horizontal direction is 22 cm and 35 cm in the up direc-
represents the positioning error comparison for the three tion. Tis positioning solution is certainly an improve-
positioning solutions. Te internal solution performs the ment from the SPP solution, which has an RMS of 2.0 m
worst in the circular track, shown in Fig. 17, due to fre- in the horizontal direction and 3.0 m in the vertical direc-
quent switching on and of of the position tracking sys- tion, in low-multipath and open-sky static environments.
tem. During this circular portion of the track, the PPP Using developed measurement prediction and stochas-
solution outperforms the RTK and internal solution. Te tic modeling approaches to condition the smartphone
internal solution failed to follow the circular track, while GNSS measurement, there is a 64% decrease in the hori-
the RMS of the RTK solution is 6.1 m compared to the zontal positioning RMS error and 100% positioning solu-
PPP solution of 5.6 m. Te RTK solution over the circular tion availability with the car dashboard data collected
portion of the track shows drifts from the reference track in suburban environments. Tere is a 62% decrease in
which is due to frequent loss of solution as the satellite the positioning error and a 23% increase in positioning
count drops to below the minimum required as a result of availability over regions where signals are signifcantly
the rejection of satellites due to missing measurements. impacted by multipath and blockages. With the help of
measurement prediction and C/N -based stochastic
0
modeling, the post-processed PPP solution had 100%
availability and outperformed RTK in terms of availability
of the solution. For the data processed, PPP is 84% more

Fig. 18 Comparison of 2D RMS error for PPP, RTK and internal positioning solution for Xiaomi MI 8, DOY 85, 2019