864
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part B5. Beijing 2008
Figure 4. Personal navigator: 1) calibration mode data flow is
shown in solid lines; and 2) when GPS signals are not available,
the dotted lines become solid indicating that now the navigation
solution is formed based on calibrated data of dead reckoning
sensors, including the human locomotion model parameter, step
length (SL) and step direction (SD).
3. PERFORMANCE ASSESSMENT
The personal navigator in the hybrid and DR modes has been
extensively tested using various operators, different terrain
configuration and mixed outdoor-indoor environments. The
details of the performance test to date were provided in
(Grejner-Brzezinska et al., 2006a-c; 2007a and b; Moafipoor
2007a and b; Toth et al., 2007), and only summary statistics are
presented here, with the emphasis on the newest results of the
mixed indoor-outdoor setting.
In this experiment, data were collected in the parking lot and
inside the Center for Mapping building on August 21 and 26,
2007. The operators, S and E, walked the parking lot and
hallways of the Center for Mapping, and made several loops
following the marked control points in the hallways of this
single-storey building. A floor plan of the building was
previously acquired by classical surveying methods, and control
points were established in the hallways with the accuracy better
than 1-2 cm in E and N, and 5 mm in height. The main objective
of the control points was to facilitate the prediction of the user’s
position and provide control for the reference trajectory inside
the building where no GPS was available. By the time the
operators started walking inside the building, they had
completed outside calibration procedures during normal GPS
signal availability (using DD carrier phase and pseudorange
measurements), which was required for a better performance of
the other sensors (see Figure 5).
Inside the building, the heading was estimated from the
HMR3000 magnetometer compass and HG1700 gyro. The
altitude was measured by the PTB220 barometer, which was
calibrated against the known pressure standards (pressure,
temperature, etc.) for the general area of activity. It was
observed that after completing the initial calibration, these
sensors showed performance that ensured redundant and
complementary measurement inputs, as well as sufficient
stability along the trajectories studied here.
Figure 5a. Center for Mapping floor plan and DR trajectory
reconstruction for operator S using compass heading.
Figure 5b. Center for Mapping floor plan and DR trajectory
reconstruction for operator E using compass heading.
Tables 7 and 8 show the accuracy assessment of the indoor DR
trajectory for one full loop along the Center for Mapping
hallways, and Table 9 provides the statistics of the three
complete indoor loops. This test represents the combination of
outdoor and indoor environments; 350 s of outdoor sensor
calibration was followed by three complete indoor loops in
three minutes (Aug. 26 dataset), using gyro/compass heading.
As can be seen in Table 9, three indoor loops are still viable
within the 3-5 m CEP50 constraint.
Test
data set
SL
model
Mean
[m]
Std
[m]
Max
[m]
End
Misclosure
[m]
CEP
(50%)
M
Operator
S
Fuzzy
0.78
0.87
1.61
2.18
0.49
ANN
1.24
0.75
1.88
2.14
1.17
Operator
E
Fuzzy
0.84
0.81
1.95
2.75
0.73
ANN
0.80
0.56
1.45
1.94
0.77
Table 7. Statistical fit to reference trajectory of the indoor DR
trajectories generated using SL predicted with fuzzy logic and
ANN, and compass heading; one indoor loop.