Page 106 - 卫星导航2021年第1-2合期

Shinghal and Bisnath Satell Navig (2021) 2:10 Page 9 of 17

Fig. 9 PPP 2D and 3D RMS positioning accuracy using three diferent stochastic models for Xiaomi MI 8 and SwiftNav Piksi measurements, DOY
146, 2019

Table 4 PPP positioning accuracy using three diferent stochastic models for Xiaomi MI 8 and SwiftNav Piksi measurements, DOY 146,
2019
Scenario Positioning accuracy Convergence
time (min)
Horizontal (m) 3D (m)
Static stochastic modeling—Xiaomi MI 8 0.81 1.35 75
C/N -based stochastic modeling—Xiaomi MI 8 0.72 1.27 58
0
Elevation-based stochastic modeling—Xiaomi MI 8 0.94 1.74 72
Elevation-based stochastic modeling—Piksi 0.14 0.16 9

weighting assignments: static, elevation-based and C/
N -based.
0
Te horizontal and vertical positioning accuracy after
convergence is presented in Table 4.
To obtain sub-metre level positioning accuracy, the
convergence threshold was chosen to be 1 m, and with
C/N -based stochastic modeling, the dataset position-
0
ing converges in 58 min, as compared to 72 min for the
elevation-based model and 75 min for the static stochas-
tic model. Te C/N -based stochastically modeled data
0
show improved initialization, convergence and least 2D
and 3D RMS error.
To validate the reproducibility of the results and the
efectiveness of the C/N -based technique over the ele-
0
vation-based weighting technique, the frst hour of data
processing for the mannequin dataset was divided into
segments of 20 min each. Terefore, the PPP processor
resets every 20 min and processes the data using the C/
N and elevation-based weighting strategy. Te horizon-
0
tal RMS of the two solutions is compared and depicted Fig. 10 PPP horizontal positioning accuracy using diferent
in Fig. 10. stochastic models for smartphone reset experiment