Installing an INS on a drone comes with a myriad of hurdles. This guide will list several pointers for improving drone setups to get the best data. The list is by no means exhaustive.
Antennas should be mounted rigidly with the INS. This means that both move as the other. It should never be the case that an antenna move without the INS also moving; there must be no independence of movement between them. Because of this, you cannot mount an INS on a gimble mount with the antennas on the top of the drone, for example.
Single or dual antenna
You can obtain good drone data performance with a single antenna however it is recommended that you install two antennas. A dual antenna setup has the advantage of receiving heading updates when the drone is stationary. This is not possible with a single antenna. Using the GNSS positions of the antennas a heading baseline can be determined. Since drones are often stationary this is important, but not necessary. If you do decide on a dual antenna setup, ensure the baseline is as long as possible. The accuracy of these heading updates is directly proportional to the baseline distance. Since space is limited on the top of a drone, a boom or prop is suggested on the underside, with antennas on either end, increasing the potential baseline.
Here is a peer-reviewed paper on the difference between single and dual antenna setups for LiDAR drone data collections.
Good navigation performance depends on precise orientation measurements. Strong orientation performance requires a good heading initialisation. In the Environment tab of NAVconfig you can configure a unit to initialise its heading via either a static or a dynamic initialisation, and you can select the speed at which to initialise. Fortunately, these are things can be changed in post-processing. An added advantage of dual antenna setups is that they can statically initialise. This is a great benefit because, due to the nature of drone manoeuvers, a dynamic initialisation can be hard to come by.
See the Flight section below for more on dynamic initialisation. If you are looking to do a static initialisation then ensure that you have as long baseline as possible. After powering on the unit static initialisation will begin immediately. Ensure you are not standing over the drone and blocking out satellite signals. There should be no obstructions around to obstruct antenna signals, nor reflective surfaces for multi-path error to occur off of. Leave the drone for two-three minutes for static initialisation to occur. This should be done at the start and end of the survey. If possible, you should connect to the unit and watch in NAVdisplay for the it to initialise before you take-off. A good static initialisation will not take long for a lock to be acquired. You should see the ambiguities reduce to 1 in NAVdisplay.
Large drones can be a source of very high Electromagnetic Interference. This is easy to underappreciate since it is not visible but steps should be taken to protect onboard instruments.
It is recommended that you install the antennas on top of ground planes for your drone. These should be circular sheets of metal, about 15-25cm in diameter. Place the antennas on top of these. This will shield the antennas from erroneous signals from below, ensuring the GNSS is as good as possible.
You may wish to use different antennas. For drone use there are many helical antennas now to choose from. OxTS units will typically be issued with Tallysman 7972 antennas, but others can be used. OxTS is only able to give specifications for our units with antennas they have been tested with.
Cables are not immune from interference and if left unprotected will suffer from signal degradation. To ensure your GNSS performance is perfect you should use cables that have certified shielding. Moreover, using shorter cables that do not have excess length (and therefore have to coiled up) is beneficial.
If your device is unable to obtain a solid Integer GNSS lock then it is quite possible it is suffering from EMI. You can test this by opening your NCOM data in NAVgraph and displaying GPS position mode. If the device struggles to get RTK Integer (6) or gxInteger (24) in a stable way in good GNSS conditions then this could be a symptom of EMI.
Check the heading, pitch and roll performance of the device throughout the flight. You can do this this by viewing heading, pitch and toll accuracy in NAVgraph (these are estimates). Or export heading, pitch and roll measurements to a CSV and calculate standard deviation. If the results are reasonably close to the specification datasheet values but the device is struggling with GPS position lock, then EMI is likely.
INS devices can cope with some noise on the accelerometers and gyroscopes and still produce an accurate trajectory. However the motors on drones produce very large vibration profiles that will cause INS devices to struggle. Steps should be taken to minimise the transfer of vibrational noise from the drone to the INS.
Rubber washers and sheets, anti-vibrational materials and platforms, as well as built-for-purpose mounting platforms are all potential solutions.
Ensure cables are securely pinned down to reduce effects on signals from mechanical vibrations.
After collecting data you can evaluate the effect of vibrational noise on the performance. Opening your processed NCOM data in NAVgraph you can view every parameter on a plotted graph. Viewing x, y, z, accelerations and gyroscope measurements you can see the variation. You can export these parameters to a CSV and calculate an RMS and a mean.
If you suspect vibrations are preventing decent performance, you can change how the processing engine deals with them. In the Environment tab of NAVconfig you can set vibrations to high or very high depending on your setup. This will get the best performance possible out of the data you have collected, however the performance will not be as good as minimising vibrations in the setup itself.
For boardset use, all of these considerations are even more important. A Boardset such as the OxTS xRED does not have a housing. As such it is more vulnerable to environmental factors like vibrations and EMI. More care therefore has to be made when mounting a boardset.
Additionally, ensure that the mounting does not flex the board. This could happen if you have screwed it on too tightly, for example.
Alternatively, you might be designing your own housing for the xRED to go into. For your own housing, ensure you have considered all of the points mentioned in this article. When evaluating the xRED, consider that housing the unit will remove many issues that are present when using a boardset unhoused.
As mentioned in the Antennas section above, a good initialisation is essential for good data. If you are set to dynamically initialise then you need to ensure a straight line of movement up to the threshold speed after taking off. The first time you hit the threshold, which is 5m/s by default, the unit will initialise its heading.
Best practice is to do a straight line for long enough for the unit to get a good lock on its heading. If viewing in NAVdisplay, you will see that the device will display 'Ready to initialise' until you hit the threshold speed, after which it will display 'locking on', and then when a lock is obtained the device will be initialised and display 'Real time'. The heading should not change during this time. Therefore do not turn the drone until it has had ample time to lock on. Furthermore the trajectory should be a straight line. This means you need to take crosswinds into consideration. It is likely you won't be able to watch NAVdisplay while the device is in flight but these principles should be borne in mind.
Try to do this at the start and the end of your data collection.
Before entering into your data collection proper, try to do as much of a warmup as possible. Excite the accelerometers and gyroscopes in each direction with manoeuvres typical of your survey flight. Figures of eights, circles, straights, and boxes are all good for getting the device warmed-up. This should not take long. It is recommended that you do a warm up flight before any data collection flight. If you are keeping the device mounted to the drone between flights then you can ensure best performance by doing a flight and improving the configuration (tightening the accuracies) with that data. Subsequent data then will not need a warm up. If the device is removed and remounted to the drone then a new warmup flight is recommended. However this is not necessary and good performance can be obtained with a small in-flight warmup.