Marvelmind Robotics forum

Hi all,

Here are the params and the firmware https://drive.google.com/file/d/0B7tPlr ... sp=sharing that we used at the Interpack.

We drew a lot of attention: https://youtu.be/M8XFkSgka8Q


Here is a flight with the Marvelmind map in the background.
https://youtu.be/YKu-Q1f48pc
As you can see the result is very good. (straight lines in a 10cm window)
You also see Stefan, the operator. He can take over control at any time or shut engines down at a flip of a switch.

Problems we encountered:
Submaps not correctly overlapping: We could not get it right. When going from one submap to the other the drone jumped.

Power supply of beacons and hedgehog. With USB connected, sometimes a beacon has a measured voltage lower than 4 volt and shuts down for a moment resulting in false position data. We soldered USB cables directly to bat+ and bat- and everything is fine.

Using different settings when building map and when flying.
During building map we let microphones of beacons send to each other (mic1 and 3) and have setting for amplifying at 2000 and threshold of 400.
During flying we use only microphone 4 on beacons and 1,2,3,5 on hedgehog with amplifying 1450 and threshold 300. 1450 because of motor interference which otherwise gets amplified too much.

It is not visible on the picture, but we ended up using 4 beacons per submap. (picture will follow)
This was because of the very small space we had to fly in.

Compass is a problem (not a Marvelmind problem). We modified the firmware to use only accels and gyros.

Height is a problem. When going up towards beacons, which are set at 185cm, you would expect that at zero angle the positioning would report 185cm, but not. It could report 0cm. Giving a lot of damage.
We set beacons at 230cm physically, and setting at 185cm. Our hedgehog is some 50cm from the ground. And we never fly higher than 150cm.

We had a successful event with a drone flying autonomously in a 2m wide isle, using Marvelmind and Pixhawk.
Marvelmind is a very good, accurate, positioning system for this application.

Some improvement can be made, like height, overlapping submaps, moving from 433Mhz to UWB, saving settings per beacon, having building and flying parameter sets, solve heading problems and full integration in Pixhawk (Ardupilot).

I like to thank Marvelmind for having a good solution and the fast responses we got when we had questions.

Hendrik

(Modified: firmware link added)

Great job hendrik. Could you share the Google drive or Dropbox address. It could be much easier to share.

Thank you for firmware. You save my graduation!!

Hey Hendrik,

Congrats on the successful flight and show! ^^
Could you post the final changes you made to the code? I need a rover FW thats why! I made the same changes you had done with using the predicted EKF and not the measured EKF. Is there any other changes?

I have completed the whole implementation of the Beacon class on ardupilot but we still having problems with the EKF! Getting a lot of compass variance which is making the position jump around everywhere! I have asked the guy who is incharge of the EKF code on ardupilot but he seems to have no time to help us out
EasyAerial guys have also tried the beacon code and reported the same problems!

EDIT: You can follow the whole conversation on this PR here!
https://github.com/ArduPilot/ardupilot/pull/6118

Hello hendrik.
I am try to applying your firmware.
There are some problems.
First I yaw value is continuously moving. Prearm check yaw value feels strict.
Second GPS North direction is moving. I don't know why maybe it is beacon's problem.
I think second is temporary phenoman but first this is really irritating. Do you have any solution?

Hi All,

The way we work with it is as follows:

In mission planner
Load firmware in PIxhawk
Set EK3_enable to 1, set EK2_enable to 0
Reboot (it will load the default EK3 params)
Load parameters from file
Reboot!
Callibrate Accels
Reboot!

In Marvelmind:
Ensure north pointing is north pointing of map and of device (drone/rover) Use button A to align, make sure compass needle B is Up

In mission planner:
Check yaw
Enter -yaw as declination at compass settings
Reboot!

Depending on accels calibration yaw moves a little.

Our experience is that accels eventually get more silent and yaw keeps pretty steady.
On the fair we had some 200 flights a day en sometimes had to reset.
Firmware accepts only yaw of +/- 5 degrees to arm.

Please look carefully at EK3 settings. As said earlier it is in a way that influence of GPS is more than Barometer. And GPS over accels for position.
On the other hand accels over compass.

Succes you all, Hendrik


https://drive.google.com/file/d/0B7tPlr ... sp=sharing

Thank you for posting the firmware and your notes about the implementation. I'm glad your product demo went well.

My application is setting up marvelmind indoor GPS for a high school robotics team to work with the Pixhawk. The kids are writing SW to direct the UAV. It is pretty much identical to your flying in a confined space.


I have a few questions.

When you say "Using different settings when building map and when flying.
During building map we let microphones of beacons send to each other (mic1 and 3) and have setting for amplifying at 2000 and threshold of 400.
During flying we use only microphone 4 on beacons and 1,2,3,5 on hedgehog with amplifying 1450 and threshold 300. 1450 because of motor interference which otherwise gets amplified too much."

Is amplifying the "Receiver Amplifier" Value ? Is threshold the "Minimum Threshold" value?

Why would you want to create overlapping submaps?

Is amplifying the "Receiver Amplifier" Value ? Is threshold the "Minimum Threshold" value?

Yes

Why would you want to create overlapping submaps?

In our situation we had two isles of glass in a L shape. The beacons on one far end couldn't hear beacons on the other far end in a straight line because of the glass, so we had to have two submaps.

Hi Hendrik!

Awesome work, really! breaking my head to do the same for the past few months!
Could you post the code so i can do the same and build a Rover Firmware? I am not able to test yours since i am trying it on rover and not copter!