1 - General context
Urban environments pose a challenge for Inertial Navigation Systems as they typically force the INS to operate in challenged-GNSS conditions, summarised in two main aspects:
- Buildings and trees obstruct the GNSS signals, leaving fewer satellites available to be tracked.
- Buildings and trees deflect the GNSS signals, causing multipath issues.
In these conditions, a larger availability of GNSS satellites can make a significant impact on the overall performance of the INS, this is why the ability of the INS to track multiple GNSS constellations can be important.
2 - OxTS units position modes
OxTS units can operate in several different position modes. Four position modes are relevant for this case study, these are listed below, in increasing order of position accuracy:
- SPS position mode: position accuracy above 50 cm. Requires L1 frequency GNSS signals.
- Differential position mode: position accuracy roughly between 20 and 50 cm. Requires L1 frequency GNSS signals and differential corrections.
- Float position mode: position accuracy roughly between 10 and 20 cm. Requires L1 frequency GNSS signals and differential corrections.
- Integer position mode: position accuracy below 10 cm. Requires L1\L2 frequency GNSS signals and differential corrections.
The most desirable position mode is obviously "Integer". This mode is not always possible to achieve, depending on the availability of GNSS signals and differential corrections.
A key aspect is that at least 6 satellites are required to achieve "Integer" position mode, in addition to a solid flow of differential corrections. Tracking more constellations gives a better chance of achieving the required number of satellites to maintain "Integer" mode for a longer period of time.
Use case: GPS+GLONASS vs GPS only
OxTS units can operate in tight coupling mode, meaning that the raw GNSS data is blended with the raw IMU data to calculate the navigation solution. this mode is referred to as gx/ix mode, it is recommended for use in challenged-GNSS environments, and has been utilised for the study case described in this article.
A dataset of about 5 hours has been collected using an RT3003 unit in the city of Torino, Italy, in a heavily urban environment, with high volume of traffic, plenty of tall buildings and trees and some short tunnels:
NTRIP real-time differential corrections, provided by the "Servizio di posizionamento Interregionale GNSS", have been utilised to improve the position accuracy of the system.
The dataset has been processed twice:
- A first time enabling the tracking of both GPS and GLONASS constellations.
- A second time disabling the tracking of GLONASS satellites
The percentages of time that the system operated in each position mode are here reported for both cases:
|GPS + GLONASS||GPS only|
|Integer mode||63.6 %||50.9 %|
|Float mode||12.1 %||5.9 %|
|Differential mode||22.3 %||41.2 %|
|SPS mode||0 %||0 %|
|none||2 %||2 %|
- As expected, the larger availability of satellites allowed the system to increase the proportion of time it operated in "Integer" mode by a significant 12.7 %. The combined "Integer" and "Float" increase was by 18.9 %.
- The 0 % of time the system spent in SPS mode indicates the differential corrections flow was solid and maintained for all of the test time.
- The 2 % of time where no position mode was reported corresponds to periods of time where the vehicle was inside a tunnel, in complete GNSS outage.
The average Northing, Easting and vertical position accuracy (RMS) are here reported for both cases:
|GPS + GLONASS||GPS only|
|Northing average position accuracy||18.5 cm||32.5 cm|
|Easting average position accuracy||19 cm||32 cm|
|Vertical average position accuracy||16.5 cm||36 cm|
- Longer periods of time on higher position accuracy modes, that is, "Integer" and "Float" rather than "Differential", translate into significantly better position accuracy on the ground.
- The reported position accuracy is an average of the entire survey, including all modes. The average position accuracy when in "Integer" mode has been around 3.5 cm, and when in "Float" mode has been around 12 cm.
The following graph and picture provide good examples of the difference in performance between the two considered cases:
The graph shows the position mode (solid lines) and position accuracy (dotted lines) of the system when using GPS+GLONASS (red) and when using GPS only (blue):
- In the first case, position mode is solidly at gxInteger(24) level, and position accuracy is accordingly around 1-2 cm.
- In the second case, position mode is only at gxDifferential(22) level, and position accuracy is around 35 cm most of the time. Only towards the end the position mode goes to gxInteger(24) and position accuracy consequently drops to 1-2 cm.
The picture is taken from Google Earth and it shows an example of situation on the ground:
- The red trace (GPS+GLONASS) shows solid and smooth trajectory, despite the challenging environment.
- The blue trace (GPS only) shows an uncertain trajectory, with the system struggling to identify its correct position whilst surrounded by buildings and trees.