Inhalt des Dokuments
Cycling Test
 UBlox EVK installed on the bike
 [1]
 © Antonoglou
This test was performed in order to check the
behaviour of the receiver in dynamic mode. For this reason the
receiver was installed on a bike, while the bike was cycling
measurements were taken. The kinematic solution of the position was
calculated in post processing mode. The measurements were taken on
29.10.2016 and there were applied three different methods to estimate
the position of the receiver, the first was the absolute positioning,
the second was the relative positioning using as base station in GFZ
Potsdam and a virtual reference station (VRS) and the third was PPP.
The first problem that occurred was that the receiver should remain
stable in the beginning for some minutes in order to be able to solve
the ambiguities. The time interval that was chosen was 510 minutes,
depending on the distance to the reference station. Additionally were
chosen two places for the cycling test. The first is Volkspark in
Potsdam, which is 8 km away from the reference station in GFZ and the
second is Tempelhof Park in Berlin, which is located 27 km away from
GFZ. These places were not chosen randomly, but on purpose. The
landscape is suitable for GNSS measurements because there are not many
trees and no buildings and the reception of the signal would not be
interrupted. Especially Tempelhof Park is ideal for such measurements
because there is nothing to block the GNSS signal. On the other hand,
Volkspark has some trees, but the landscape is generally good. The
combination of the dynamic motion of the receiver with the landscape
will simulate the measurements on a flying UAV.
The change of
position was calculated in respect with a point wich is located on the
center of gravity of the measurements in each case. The positions of
that points were used later as positions of the VRS.
Results
Absolute positioning
 Volkspark (position) [2]
 Volkspark (standard deviation) [3]
 Tempelhof (position) [4]
 Tempelhof (standard deviation) [5]
Relative positioning (base station in GFZ)
 Volkspark (position) [6]
 Volkspark (standard deviation) [7]
 Tempelhof (position) [8]
 Tempelhof (standard deviation) [9]
Relative positioning (base station VRS)
 Volkspark (position) [10]
 Volkspark (standard deviation) [11]
 Tempelhof (position) [12]
 Tempelhof (standard deviation) [13]
PPP
 Volkspark (position) [14]
 Volkspark (standard deviation) [15]
 Tempelhof (position) [16]
 Tempelhof (standard deviation) [17]
Examining the results from the absolute positioning, there are many irregularities appearing very often in the trajectories and they are not smooth generally. There is no clear evidence that those changes are highly correlated with the changes in the standard deviation of each measurement because of their frequency, especially in Volkspark. The biggest problem is in the height component. Independently from the positioning method, the heights are more erroneus in GNSS because of the geometry of the satellites. If there was possibility of observing satellites bellow the horizon, the errors would be equally distributed in all dimensions. In Tempelhof the quality of the measurements is better, the reasons of this fact are mainly the multipath effect and the slip circles. The landscape in Volkspark is not as good as in Tempelhof, there are many multipath effects and the trees are blocking temporally the signal, this problem is almost vanished in Templehof. In both cases the standard deviations are generally linear, but with many irregular ”jumps”. That ”jumps” occur due to an appearance of a new satellite (or loss of one) and/or due to the slip circles. The precision is generally low in absolute positioning (especially in kinematic mode), additionally the effects described above, result in an instability.
RMS σ_{e }(m)  RMS σ_{n
}(m)  RMS σ_{u
}(m)  

Volkspark  1.298  1.837  4.114 
Tempelhof  1.236  1.492  3.493 
Relative positioning can give the best positioning solution. The results from the 1^{st} solution (base station in GFZ) show that the trajectories are generally smooth, but there are some irregular and rapid changes in the position. Moreover, the noise of the heights is signifigantly lower in this case. The rapid changes are highly correlated with the changes in the standard deviation of each measurement. The precision of the position depends on the distance to the reference station, the landscape and the time of the measurement. Even though Tempelhof is further away from Volkspark, in respect to the base station, the errors are smaller and smoother along the time. The reason of this fact are the same parameters that destroy the pressision in absolute positioning (multipath effect, slip circles and an appearance or loss of a satellite). After any “jump” the precision is improving usually along the time, if there is not any additional intermediate “jump”. In the case that was used a doublefrequency receiver, this problem would not appear so often and the precision would be stabilized faster. The magnitude and the repeatability of these effects change among the time, even if the measurements are taken in the same place. This phenomenon happens due to the change of the geometry of the satellites. In Tempelhof, during most of the time of the measurements, the standard deviation has the same behavior. This occurs because the multipath effects and the temporal losses of the signal are reduced significantly. An other reason for the rapid changes in the heights is the motion of the rider. That motions can be clearly shown in the graphs.
RMS
σ_{e }(cm)  RMS
σ_{n }(cm)  RMS σ_{u
}(cm)  

Volkspark  6.96  8.58  20.96 
Tempelhof  6.95  8.36  17.69

An alternative solution was given by using a VRS as a base station, instead of the thestation that is located in GFZ Potsdam. Comparing the trajectory with the previous, there is no significant difference. The standard deviations have different shape, but similar behaviour. This was expected because the algorithm for the solution is the same in both cases. Comparing the tables with the RMS, the precision is slightly improved in Tempelhof and in Volkspark it is slightly decreased. Theoreticaly the results should be better, but there is no big difference. The reason that lead to this fact is that in every case are used different sets of satellites. Furthermore, the number of the common satellites is reduced in the case of the VRS. This fact is critical and the solution is not improved.
RMS
σ_{e }(cm)  RMS
σ_{n }(cm)  RMS σ_{u
}(cm)  

Volkspark  7.13  8.84  21.41 
Tempelhof  6.67  7.98  17.20

A final solution was given by PPP method, PPP is a special case of absolute positioning because it requires one GNSS receiver. The trajectories are not as smooth as those from the solution using relative positioning, but the irregular changes in the position are observed on the same epochs in both solutions. Similarly to the previous solution, the precision in Tempelhof is better, but the big difference comparing both methods is the behaviour of the standard deviation. There is a correlation between the rapid changes,the ”jumps”, of both solutions, but the standard deviation in PPP is not smooth. For this reason there is no safe conclusion about the appearance of slip circles, multipath effects or any other phenomena that reduce the accuracy. This difference appears because both methods use a completely different strategy to find a positioning solution.
RMS σ_{e }(cm)  RMS σ_{n
}(cm)  RMS σ_{u
}(cm)  

Volkspark  27.81  30.79  89.69 
Tempelhof  26.99  30.88  77.72 
bike.png
_sin_pos_2.png
_sin_std_2.png
sin_pos_2.png
sin_std_2.png
_rel_pos_2.png
_rel_std_2.png
rel_pos_2.png
rel_std_2.png
s_vrs_pos_2.png
s_vrs_std_2.png
_vrs_pos_2.png
_vrs_std_2.png
s_ppp_pos_2.png
s_ppp_std_2.png
_ppp_pos_2.png
_ppp_std_2.png