SMARTNAV RTK Test – GUIDE GNSSJuly 28, 2016 3:25 pm
SMARTNAV RTK Test with GUIDE GNSS in urban areas
Drotek, based in Toulouse (France) has had the opportunity to test SMARTNAV L1 RTK with laboratory grade devices. This test has been run in straight collaboration with GUIDE (GNSS Usage Innovation and Development of Excellence), a testing laboratory for satellite geolocation.
It is nowadays quite clear that L1 RTK performs really well in “easy” open sky environnements. But we wanted to test its performances in really challenging environments.
GUIDE is equipped with a GBOX, a “black box” containing an aeronautical grade Inertial Measurement Unit and a multiconstellation L1/L2 AsterX receiver.
The antenna was placed on a car’s roof.
We set up two devices on the car :
– one standalone SMARTNAV (with its own antenna), logging raw data for post-processing,
– one SMARTNAV connected to the car’s antenna splitter, processing real-time NRTK (VRS) with Teria network corrections.
The environments we wanted to test were the following :
– Height differences (>10%)
– Bridges and tunnels
– Large round abouts
– Housing estate
– Urban canyon
– Leafy streets
The following picture displays two trajectories :
- – The blue trajectory is the reference trajectory,
- – the other is L1 NRTK pseudo-raw output (red is for single solution, yellow is for float solution and green for fix solution). It is a pseudo-raw solution because RTK double differences algorithm is applied all the time with no sanity check (we will see later that almost all position jumps can be corrected with a sanity check). Single positioning occurs when no differential data is available, or when signal strength is too low. It is the same output as a L1 classical GNSS. Float solution occurs when ambiguities are not fixed (the integer number of wavelengths between the satellites and the receiver are calculated with a float approach). Fix solution occur when the ambiguities are fixed (this is when maximum precision is reached).
1 – Starting point
It is important to compare the two outputs at the beginning of the test, when the vehicle has been standing still for a while.
L1 NRTK takes as expected several minutes to converge (approximately 7 min with a covered sky view, one masking building). Strangely, reference position convergence is not immediate, even though it is really fast. There is a position bias of 70 cm between the two. The car remained static for 30 min once converged. During this time there was no point outside the converged scatter plots.
We can already see that the L1 NRTK’s spreading is not really different from reference, suggesting a future hypothesis that L1/L2 systems are not more precise but only more available than L1 systems.
On the overall, L1 NRTK got 27% fix solutions, 62.5% float solutions and 10.5% single solutions. A post-processing on the L1/L2 raw data (without IMU hybridation) shows 50.6% fix solutions, 46.1% float solutions and 3.2% single solutions with the exact same data. Care must be taken when interpreting these numbers because post-processing does take into account that differential data transmission can be delayed or even erratic, especially in urban dense areas.
2 – Getting on the road
The same offset is present during the first part of the test. It is precisely constant.
The first trees disturb signal. Highest position discrepancy with reference trajectory is about 2m.
3 – Height differences
Whenever a fix occurred during the test, the maximum deviation observed between trajectories was about 20 cm. However, there was no LTE signal during this part of the test, so the system’s output was a single solution.
The output is U-Blox M8T direct output, with the dynamic model filter. The reliability of the solution is quite good, deviation oscillates between 0 and 5 m around reference trajectory.
As soon as differential data is back reacquisition is fast even if there is a submeter offset at the beginning.
4 – Bridges and tunnels
On the road with good sky view the deviation between two trajectories is better than 20 cm, even with float solution.
The purpose of this test was to test how fast signal reacquisition is done and how it affects solution quality.
Reacquisition is almost instantaneous and solution convergence only takes a few epochs.
5 – Large round abouts
Deviation with reference trajectory is excellent, even if it is difficult to measure RTK repeatability with these measurements.
6 – Housing estate
This environment becomes really challenging for NRTK. On the overall trajectory follows reference trajectory but there area high number of jumps in the raw NRTK output.
7 – Urban canyon
With the smallest sky view possible, the output is not reliable at all.
Even the reference system does not have any output at some point.
8 – Leafy streets
Leaves absorb, reflect signals and create a high quantity of multipath. At the moment of the test there was almost no direct sky view. This results in lots of jumps and an unreliable position.
9 – Repeatability
The urban test did not let us measure the repeatability of the system, so we ran an additional test : we tried to draw with the RTK output the border of a square table :
The path was repeated several times and the measured deviation did not exceed 5 cm.
10 – Insights :
- – L1 NRTK shows good results in moderately obstructed sky views
- – L1 NRTK is not reliable in heavily obstructed sky view, but fused with L1 single it can still be highly reliable thanks to M8T sensitivity
- – Good repeatability is achieved with this system
- – Accuracy depends on time of convergence
- – Reacquisitions are fast
- – At some points some offset appear, but it is difficult to say where they come from. A post-processing on the L1/L2 data (without IMU hybridation) also shows similar offsets from reference trajectory. Could the logging technique (interruptions between each test) have worsened the quality of the reference trajectory?