5
the dates were random effects. In the model with data from
all dates combined, an autoregressive correlation matrix
was used to model the correlation between the repeated
measures at the 10 reflectors. The autoregressive correla-
tion assumes pairs of measurements closer together in time
will be more highly correlated than pairs of measurements
farther apart in time. For this analysis, the P-values were
less than 0.05 and therefore considered to be statistically
significant. The outcomes for the static and dynamic data
analyses are discussed in the Results section below.
RESULTS
Static Test Results
Figure 3 shows a comparison of the coordinates measured
by the two different systems for the 20 points on the 10
reflectors. It can be found seen in Figure 3 that the coor-
dinates measured by the GNSS system agree well with the
coordinates measured by the RTS system. Figure 4 shows
the difference (i.e., the error shown in Figure 4) between
the coordinates measured by the two different systems for
all 10 reflectors. The average error in the X coordinates was
less than 0.5 cm, and the average error in the Y coordinates
was less than 0.8 cm. The median errors in the X and Y
coordinates were less than 0.5 cm and less than 0.8 cm,
respectively.
Dynamic Test Results
The results showed the estimated distance error in posi-
tional accuracy of the GNSS receivers for all variables such
as day of testing, GNSS status, direction of travel (forward
or return/reverse), and vehicles, independently. As shown
in Table 1, NIOSH researchers estimated a distance error of
1.34 m (4.40 ft) with a 95% confidence interval (0.99 m,
1.69 m) (3.2 4 ft, 5.54 ft) using RTK and 1.50 m (4.92 ft)
with a 95% confidence interval (1.15 m, 1.85 m) (3.77 ft,
6.07 ft) using SBAS. These error estimates were indepen-
dent of vehicles, the different date of testing, the directions
of travel, and the different speed. Independent of the date
of testing, GNSS status, speed, and vehicles we estimated
distance errors of 1.09 m (3.57 ft) in the forward direc-
tion and 1.75 m (5.74 ft) in the reverse or return direction.
Independent of GNSS status and date of testing direction
of travel, our model yielded results for all the speed catego-
ries of both vehicles. For the UGV, we estimated a distance
error of 1.38 m (4.52 ft) at 1.27 mph or low speed (L) and
a distance error of 1.41 m (4.63 ft) at 2.59 mph or high
speed (H). For the e-truck, we estimated distance errors of
1.34 m (4.5 ft) for 5-mph tests, 1.43 m (4.69 ft) for the
10-mph tests, and 1.55 (5.09 ft) for 15-mph tests.
DISCUSSION
Independent of the date of testing, GNSS status only influ-
enced the distance errors for tests conducted using a UGV
and not the results for the e-truck. For the UGV, the dis-
tance errors in positional accuracy were less for tests that
were run with RTK corrections than those with SBAS—
for all variables independently and in combination (see
Table 2). Independent of speed, NIOSH researchers mea-
sured distance errors of 1.25 m (4.10 ft) moving forward
and 1.11 m (3.64 ft) moving in reverse. With SBAS, we
measured distance errors of 1.71 m (5.61 ft) moving forward
and 1.49 m (4.89 ft) moving in reverse. Independent of
direction of travel, at high speed (H), we measured distance
errors of 1.15 m (3.77 ft) using RTK and 1.66 m (5.45 ft)
Reflector 1
Reflector 10
Figure 3. A comparison of the measured coordinates for the
ten reflectors mounted on the ground using two systems: the
RTS (serving as the ground truth) and the GNSS receivers
Figure 4. Static results comparing the GNSS receivers to our
RTS-based surveying equipment
Error
(m) Error(m)
the dates were random effects. In the model with data from
all dates combined, an autoregressive correlation matrix
was used to model the correlation between the repeated
measures at the 10 reflectors. The autoregressive correla-
tion assumes pairs of measurements closer together in time
will be more highly correlated than pairs of measurements
farther apart in time. For this analysis, the P-values were
less than 0.05 and therefore considered to be statistically
significant. The outcomes for the static and dynamic data
analyses are discussed in the Results section below.
RESULTS
Static Test Results
Figure 3 shows a comparison of the coordinates measured
by the two different systems for the 20 points on the 10
reflectors. It can be found seen in Figure 3 that the coor-
dinates measured by the GNSS system agree well with the
coordinates measured by the RTS system. Figure 4 shows
the difference (i.e., the error shown in Figure 4) between
the coordinates measured by the two different systems for
all 10 reflectors. The average error in the X coordinates was
less than 0.5 cm, and the average error in the Y coordinates
was less than 0.8 cm. The median errors in the X and Y
coordinates were less than 0.5 cm and less than 0.8 cm,
respectively.
Dynamic Test Results
The results showed the estimated distance error in posi-
tional accuracy of the GNSS receivers for all variables such
as day of testing, GNSS status, direction of travel (forward
or return/reverse), and vehicles, independently. As shown
in Table 1, NIOSH researchers estimated a distance error of
1.34 m (4.40 ft) with a 95% confidence interval (0.99 m,
1.69 m) (3.2 4 ft, 5.54 ft) using RTK and 1.50 m (4.92 ft)
with a 95% confidence interval (1.15 m, 1.85 m) (3.77 ft,
6.07 ft) using SBAS. These error estimates were indepen-
dent of vehicles, the different date of testing, the directions
of travel, and the different speed. Independent of the date
of testing, GNSS status, speed, and vehicles we estimated
distance errors of 1.09 m (3.57 ft) in the forward direc-
tion and 1.75 m (5.74 ft) in the reverse or return direction.
Independent of GNSS status and date of testing direction
of travel, our model yielded results for all the speed catego-
ries of both vehicles. For the UGV, we estimated a distance
error of 1.38 m (4.52 ft) at 1.27 mph or low speed (L) and
a distance error of 1.41 m (4.63 ft) at 2.59 mph or high
speed (H). For the e-truck, we estimated distance errors of
1.34 m (4.5 ft) for 5-mph tests, 1.43 m (4.69 ft) for the
10-mph tests, and 1.55 (5.09 ft) for 15-mph tests.
DISCUSSION
Independent of the date of testing, GNSS status only influ-
enced the distance errors for tests conducted using a UGV
and not the results for the e-truck. For the UGV, the dis-
tance errors in positional accuracy were less for tests that
were run with RTK corrections than those with SBAS—
for all variables independently and in combination (see
Table 2). Independent of speed, NIOSH researchers mea-
sured distance errors of 1.25 m (4.10 ft) moving forward
and 1.11 m (3.64 ft) moving in reverse. With SBAS, we
measured distance errors of 1.71 m (5.61 ft) moving forward
and 1.49 m (4.89 ft) moving in reverse. Independent of
direction of travel, at high speed (H), we measured distance
errors of 1.15 m (3.77 ft) using RTK and 1.66 m (5.45 ft)
Reflector 1
Reflector 10
Figure 3. A comparison of the measured coordinates for the
ten reflectors mounted on the ground using two systems: the
RTS (serving as the ground truth) and the GNSS receivers
Figure 4. Static results comparing the GNSS receivers to our
RTS-based surveying equipment
Error
(m) Error(m)