Low cost L1+L2 GNSS RTK receiver for Drones. North RTKite receiver.

Darius Jack February 22, 2016 at 1:12pm

@RPM,

thank you

100km/h = 100,000 m/ 3,600 s rounded to 28 m/s

if you fly your drone slowly at 10 km/h so 1sec GPS offset calculated is 2.8 m

If your GPS is clocked at 10Hz and no latency is generated by Pixhawk code, motors control loop, don't expect better GPS accuracy than 0.28 m irrespective your GPS is RTK, L1 RTK , L1/L2 RTK

Since GPS can be clocked at 3Hz, 5Hz, we can assume 1m GPS geolocation accuracy error by default.

Drone is fast moving Rover, so geodesy grade GPS accuracy doesn't work for it.

In the worst case, as calculated by RPM, if you fly a drone at 100 km/h and your GPS is clocked at 1Hz and Pixhawk code execution latency is set to 0 miliseconds, you can still expect 28 metres GPS error.

I am not sure if this is one of the main causes of drone fly-aways but should be checked and verified..

@Bernardo,

If I recall correctly, GPS + IMU (innertial navigation) has been already embedded into Pixhawk controller.

Wish you success at Barcelona Mobile Congress.

Bernardo Espinosa February 22, 2016 at 1:06pm

Darius, here is a sample data set I've found.

The data is logged straight into the RTKite as NMEA and converted to KML.

The test was done at ground level, with the RTKite mounted on a bike, moving fast, on a park surrounded by buildings, and tall trees. Nightime, half overcast.

The log is at 10Hz, and done on strict RTK, without postprocess edition or cleanup, it is raw out of the receiver. The correction source was the freely available CatNet CORS sending GNSS RTCM3 at 1Hz.

One recording is Dynamic and the second recording is static, meaning that the receiver was not moving, but it was recording dynamic RTK at 10Hz. (You will notice a spike at the end of the recording due to the body covering the antenna while stopping the recording)

You can use Goggle street maps to see the surroundings. Urban canyon with high multipath noise.

https://onedrive.live.com/redir?resid=37E7F12A3066E006!8606&authkey=!APw8x_phbmoB6U8&ithint=file%2czip

Looking forward for your comments.

Bernardo Espinosa February 22, 2016 at 12:51pm

(I think my past comment was lost)

Darius, it seems an interesting test. I will supply you some sample data at 10Hz RTK. (we are in the midst f the Mobile World Congress)

Actually we give 10Hz for fast moving applications on both air and ground, so there's no problem.

Besides that and considering the faster fixed wing drones, and in order to geotag at 150Hz (nearly instantaneous) the internal algorithm uses the XYZ acceleration to readout the precise spatial location when the camera shoots.

RPM February 22, 2016 at 12:24pm

@Darius, is that a straight comparison from km/h to m/s? If so, 100km/hr rounds to 28m/s

Darius Jack February 22, 2016 at 12:16pm

@Bernardo,

@Anton,

I would like to support your L1 vs. L1/L2 GPS accuracy challenge at my Open FabLab.

What is offered by Bernardo is ground station + rover configuration known from geodesy.

Rover in geodesy is not mobile or is slow moving GPS rover and geolocalization is calculated at fixed point.

The opposite configuration comes with a drone.

Drone is fast moving gps rover.

GPS is clocked 1Hz - 10Hz

drone controller, motor controller are clocked by code loops at known timing.

If your drone flies at 10 km/h forward speed , so about 3 cm/sec

If 100 km/h, so 0.30 m/s

so if GPS clock is slow and controller processor is slow (not real-time)

don't expect better than 1m accuracy since latency counts and calculated fix

is always the previous fix, since drone is moving fast.

Since GPS raw data can be saved to a file and postprocessed,

just record GPS and RTK data files and let us verify claimed accuracy of L1 vs. L1/L2

technologies.

a brief note on L1, L2, L5

http://geoconnect.com.au/gps-signals-l1-l2-l5/

GPS signals (L1, L2, L5)

Bernardo Espinosa February 22, 2016 at 9:32am

Dear Anton, kindly write to emea@northgps.com to have every detail.

As said, your R+D looks promising. Cheers!

Anton Strydom February 22, 2016 at 9:23am

Good day Bernardo

I will forward you all the data I have for your perusal. I am not in R+D phase but am moving to production. The environments I have tested in is in a vehicle driving on the highways at 120 km /hr. Flying a Jabiru 4 seater plane at around 200km /hr and then in mopani veld as well as under tree canopies in the Caprivi strip in Namibia

I am not looking again to start an argument but simply indicating that L1 has been neglected over the years and that you cannot compare results from an autonomous GPS to a GPS capable of dong RTK

For what it is worth we used a Leica 500 and our GPS in the plane with both the antennas inside the plane behind the passenger seat. The performance was identical. I would be interested in getting one or two of your L1/L2 receivers and put them on my GPS to evaluate then against the L1 receiver I am using now

What would the cost of the receiver only be?

Regards

Anton

Anton Strydom February 22, 2016 at 9:08am

Good day RPM

I hear what you say,

I am not trying to get into an argument. If you look at the data posted here it is comparing a RTK capable device to a autonomous device. The U-Blox they use for comparison is not capable of doing RTK

L1 has been shoved aside with the availability of L2/L2c and L5. What we set out to do was to produce an affordable instrument capable of locating objects such as survey pegs etc due the distances my brother has to travel and the cost involved in having to travel and the point out a few pegs 9distance from Windhoek to Rundu around 800 km and to Katima around 1500 km) Besides keeping or one of his staff members out of more productive work.

It has taken me almost 3 years to develop this unit and using L1/L2 equipment to test against.

I respectfully disagree that L1 looses its fix in less ideal situations. It is purely as I state above that with the advent of L2/L2C and L5 that L1 was disregarded and no development took place.

Why it has taken me so long to develop this unit is purely that I am funding the development out of my own pocket and do not have the resources companies have.

I will with pleasure answer any questions and assist where possible

Maybe I should as you suggest create a blog for this

Take care

Regards

Anton

Bernardo Espinosa February 22, 2016 at 8:58am

Dear Anton, the comment that RPM does is the correct point. Please comment us regarding the environment of your trials.

Please note the samples of operation that I showed under 15m high tree canopy, rainy and cloudy ionized environment, that is the commonplace on the fieldwork. As you may imagine the people that create and launch L2 capable satellites wouldn't do it if it wasn't needed.

Also consider that is better that you start your own discussion as you are still on R+D phase, and you will receive very important feedback.

Looking forward for your thread and certainly interesting data!

RPM February 22, 2016 at 8:47am

Anton,

In my experience the attraction to L1/L2 over L1 is not necessarily greater accuracy. It is rather the L1/L2's ability to hold fix in less than ideal situations.

I am sure everyone involved in this thread would be interested in learning about your product through its own blog post.

Looking forward to all developments!

Blogs

Low cost L1+L2 GNSS RTK receiver for Drones. North RTKite receiver.

Comments