Dear Community,
Drotek, French tech start-up is proud to release its L1 RTK GNSS ! RTK technology allows to enhance GNSS positioning and to go down to centimeter level precision.
We have seen a lot of RTK boards coming up. However we have witnessed no good integration for the UAV segment and no performance enhancement in difficult environments.
This is why we decided to develop a solution to cope with these issues. Drotek is specialized in electronic design for UAVs and has put all its expertise in designing this board, jointly with French National Center for Space Studies (CNES) and French Institute for Aeronautics and Space (ISAE).
What is RTK for?
RTK technology uses corrections from a second GNSS receiver (called “base”, the other mobile receiver is called “rover”) in order to correct errors that lead to meter level precision. It has been used for a long time by surveyors, but at a very high cost. Today we want to trivialize this technology.
Our two years spent on developing this product have led us to focus on several points, such as connectivity, signal quality, EMI, real-time, etc...
We have compared our system’s performance with dual frequency receivers (in blue, Rover : Septentrio AsteRx 3 L1/L2 + Novatel geodesic antenna in RTK with base Trimble NetR9 + TRM59800), and it is pretty astonishing!
RTK works well in clean open-sky environments, but we like to test our device in difficult scenarios. Urban canyon is stil a pipe dream, but some route on ground robot in medium-hard environment (buildings, trees…) gives pretty good results :
Comparison between single GPS (red) and RTK : have a look at the left side under the tree. Single position gets highly contaminated by multipath, whereas the big ground plane and the Tallysman antennas filter it quite well :
Comparison between RTK Drotek (orange) and another open-source RTK solution (yellow) :
Base correction data can be sent through serial port with 3DR Radio types but also through cabled or wireless Internet for those want to get rid of range limitations. All drivers and server programs are already in the board. The following table sums up the different possible ways of communication between rover and base.
We are currently developing LoRa based transmitters, hoping that these will fulfill our expectations (we are quite excited about them!). Coming soon.
We truly believe in open-source community, but this board has also been designed to work “as is”. All boards will be delivered flashed and ready-to-use, with really little configuration, mainly for non-advanced users who would wish to get to precision without spending too much time on understanding how it works. Our wiki will explain everything step-by-step, but will also provide support on using RTKLIB graphical tools for post-processing and analyzing data. Do not forget to check our Github, we will publish in it really soon!
All our boards are stamped “Made in France”, we pay a lot of attention to the devices we manufacture in our French factory. Our motto is open source and open hardware, we work this way and we will stay this way. We want sell the atoms not the bits.
The package will be available in our site for pre-ordering. The expected release date is in one month, with a price of 699 € excluding VAT.
Our website is currently being redesigned, hope you will appreciate the new one! There will be plenty of content on how to use all the devices we manufacture, from little sensors to big boards!
Some additional exciting results :
This picture depicts the precision you can expect with good conditions, amazing!
A little demonstration (shot yesterday!) :
We modified an Iris+ drone, removed its GPS and replaced it by our board. Take-off is manual, and as soon as the drone lifted we triggered Return To Launch mode. The drone succeeded in landing on the table approximately 1 out of 2 times. It touched the table with at least two feet each time we tried the procedure. Wind conditions were quite gentle but with a fair amount of gusts. Even if the board provides precision, there has been a lot of work on the controller’s PIDs to achieve this result!
We used a Netgear WiFi router to transmit raw data from the base to the rover, using Edison's WiFi. Base was connected in Ethernet to the router, everything powered by a 4S LiPo and a 12V power module.
We will offer a board to the person who comes up with the coolest name for the device !
10% off on pre-order !
See you soon in our website !
Comments
Darius has a habit of hijacking every single blog post that's been posted in the last 6 months.
Thank you Kevin,
but if you publish a single RTK GPS accuracy test, so you exactly claim 5mm accuracy
of your RTK GPS since you don't publish other test.
5mm RTK GPS accuracy test is unreliable since 5cm is claimed by developer of RTKLib.
Alike problem is with your comparison of 2 RTK GPS units.
Just disclose the name of other manufacturer to let buyers to request more tests to be done, tests supervised by both manufacturers.
Today market of RTK GPS totally collapsed since none manufacturer claims full understanding of the theory behing RTK corrections.
Pure implementation of open source RTKLib on Linux machine makes no sense
since your oponent did exactly what you did it, implementing RTKLib alike way.
I am working on software-only high accuracy GPS technology since 70km
ground station - rover away high accuracy solution offered in case of geodesy grade rover unit is hard to believe is you study GPS Ionospheric error maps in 24h cycles
to show 5-10 metres offset for any fixed geolocation.
I am satisfied with 5-10 metres GPS high accuracy claimed by Ublex
since this is for real ( 200-page long manual and paper)
But 5mm accuracy is highly unreliable as opposed to claimed by developer of RTKLib
- 5cm accuracy.
We should get real
5cm makes 10 x 5mm
So is your test unreliable or theory behind RTKLib should be revised once again.
Hi Darius,
We do not claim 5 mm absolute accuracy, our test just shows that in a fairly short amount of time relative accuracy can be within this scale. The point that is geolocalized with that accuracy is the antenna's phase center. Corresponding to Tallysman's documentation, its variation is very tight and should correspond to antenna's center.
Thank you Kevin,
"
In the end, you can do RTK with only few satellites, the key is to find the algorithm that selects them right."
You are still required to select satellites to get HDOP and PDOP within limits.
"
If you try the same test on another day you can find the same scattering 5 cm away."
So why didn't you test your RTK GPS, marketed product for a whole month for the fixed geolocation ?
If you claim 5mm accuracy I would like to know what point within GPS hardware + antenna, is represented on your plot.
Just tell me what point within GPS hardware + antenna ist geolocalized at 5mm accuracy ?
Hi Darius,
For the moment only RTKLIB is implemented, we added the protocol to communicate with Pixhawk. We are also studying RTK algorithm itself in partnership with ISAE and CNES, for example how to deal with multipath in urban areas using doppler measure, how to get a real indicator of the solution quality (MLAMBDA does sometimes give you an ambiguity fix, but it can be a false fix)... but this is still under testing.
The 5mm accuracy you find in the picture does not mean it is an absolute precision of 5mm, it only means that in this amount of time (maybe tens of minutes) you have this relative precision. If you try the same test on another day you can find the same scattering 5 cm away.
Base's location was on the roof, third building from the right (grey one).
Darius I am sorry I do not understand your last question, are you talking about base's position? Or antenna's phase center? I will be happy to answer you if you clarify it please.
It is true that we see plenty of RTK devices in the market nowadays. We have tested both GPS only RTK and multiconstellation systems and there can be no comparison between both in my opinion. With GPS only you can experiment high offsets even with stable trajectories. In the end, you can do RTK with only few satellites, the key is to find the algorithm that selects them right.
did you study the idea behind RTK corrections or just have implemented
RTKLib ?
I am really surprised by your claimed accuracy of 5mm
if Prof. Takasu claimed 5cm : 5 mm vs. 5 cm
If you claim problems with another implementation of RTKLib
I would prefer to know the exact name and contact of third party developer of RTKLib to let us verify your claims in peer-2-peer live talks.
"
Comparison between RTK Drotek (orange) and another open-source RTK solution (yellow) :"
What was the geolocation of your ground reference station ?
I have studied GPS ionospheric error maps and to build real high-accuracy RTK GPS you need to do more to get HDOP PDOP received from ground station to match one calculated at the rover
and since M8 can support up to 5 constellations and you can select constellation, it makes sense to select the same constellation on both units.
At the same time I would like to read third party analysis of theory behind the claimed accuracy of RTK GPS.
Maybe we need more than one ground station GPS.
I would like to know, for comparison, which exactly point in antenna, GPS PCB is represented on your GPS plots ?
Could you point to this point on image of your RTK GPS ?
thanks.
BTW
I would like to test every RTK claimed GPS and test them all
in parallel.
Otherwise market gets full of $600 RTK GPS solutions vs. L1 L2 $60 GPS solutions, either claiming 5mm, 5cm or 5m high accuracy.
darius
manta103g@gmail.com
Great Project!
@Emlid
Maybe this helps motivate you to put some more effort into the reach project? I mean the software is still beta at best and the Pixhawk integration is nowhere. Also your "todo" list on your wiki looks pretty much the same as it was in the begining...
@Agribird, thank you for your words of encouragement! Hope it will help you in developing precision farming!
@Thorsten, LoRa modules will have two SMA connectors, one for 433 MHz, the other one for 868 MHz. We will perform range tests in a few days and will report them!
Hi Thorsten,
New firmware 3.01 for M8T allows you to use Galileo data. But for the moment RTKLIB does not process this data for the moment (version 2.4.3 does, but it is still in beta, available in Tomoji Takasu's Github). I have seen reports of frequencies up to 12,5Hz in GPS/GLONASS/SBAS/GALILEO with M8T.
I usually disable SBAS for testing because of the timing issues, never seen any problem at 10Hz then.
-
4
-
5
-
6
-
7
-
8
of 8 Next