Replies to This Discussion

Permalink Reply by Andrew Klofas on July 29, 2009 at 4:30pm

I've done a UAV helicopter augmentation control system and I don't think it is possible to rely only on the gyroscopes for orientation because they are too noisy. Fixed wing aircraft have only one direction (forward), so you can infer the orientation using the gyroscopes and GPS, but that is not the case for helicopters. In addition, high vibration is really hard to kill and adds *more* noise to the sensors. You might need to add a magnetometer sensor module to the heli. I'm interested in this direction cosine matrix, so you have a link to the paper?

Thanks,
Andrew

• ▶ Reply

Permalink Reply by John_Mac on July 29, 2009 at 4:35pm

Andrew,

Those are the things we intend to chase down. There is also another gyro (more expensive) that seems less sensitive to the vibration spectrum we expect to see.

All of the UAV Dev Board info I referenced is at this URL

http://www.diydrones.com/page/uav-devboard

Best,
John

• ▶ Reply

DeveloperPermalink Reply by William Premerlani on July 29, 2009 at 5:00pm

Andrew,

The combination of the hardware design that is used for the UAV DevBoard, and the direction cosine matrix algorithm (DCM) that Paul Bizard and I are using (based on the work of Robert Mahoney) to combine accelerometer, gyro, and GPS information is so quiet and noise-free, that unless you move the board, you will think that the power is off. Yet the response is fast and accurate when you rotate the board. There now is a very large experience base with a number of different airframes.

So far, all the work with the UAV DevBoard and DCM has been done has been on airplanes, so its not clear how well it might work on helis. John Mac is the first to give it a try to find out.

The biggest unknown is vibration. Others have reported a vibration issue with the type of gyros that are on the board. Early on I did some tests and thought there was a vibration problem, until I realized what was really happening was aliasing. I raised my sampling rate to 5000 samples per second and did some digital filtering, and that problem went away. So maybe vibration will not be an issue. John Mac is going to do some testing to find out. As far as I know, he is the first one to try the board on a heli.

Regarding the need for a magnetometer, we might be able to get away without one. It turns out that in the DCM algorithm, you do not need to know yaw in order to compute pitch and roll accurately. Gyros and accelerometers will do the job perfectly well. So, pitch and roll will be fine without a magnetometer.

With respect to yaw, the yaw drift is pretty low, so in a hover, although you will not know which way the heli is pointed, it will yaw very slowly. And then when you start going forward, the DCM algorithm will quickly lock onto the GPS information, so you can properly navigate.

I've done something similar with an RC truck, without a magnetometer. Its how I got started. Its fine when you're running it around. When you stop it, it looses track of north, but when you start forward again, it quickly locks back in and smoothly swings to the direction it needs to go.

Best regards,
Bill

• ▶ Reply

Permalink Reply by John_Mac on July 29, 2009 at 5:07pm

Just an additional note, I will be using the gyro on the TREX directly for yaw control at this stage of development. So the UAV Dev Board will only be providing stabilization for pitch and roll.

John

• ▶ Reply

Permalink Reply by Andrew Klofas on July 29, 2009 at 5:11pm

That sounds really neat! I would very much like to track this project and if you guys need me, help out. I live in the bay area, which may or may not be near your location. In any case, let us know your progress!

Andrew

• ▶ Reply

Permalink Reply by John_Mac on July 29, 2009 at 5:29pm

Thanks Andrew

Like I said, I hope to get more info on this Blog when I get back home next week. I see this as Stage I...have lots of aspirations:) but one step at a time. Hopefully, others can take on enhancements (telemetry is one I am interested in) and take it to the next level.

Best,
John

• ▶ Reply

Permalink Reply by Miguel COLLADOS on July 30, 2009 at 10:22am

Concerning telemetry, may be I could help:
For my FPV I am using an OSD wich sends telemetry back to me (current consuption, battery voltage, distance to home, heading, altitude, gps speed..) by te means of a 500mW video transmitter so I can see it superimposed over the camera image, but the data is available and could be sent with a modem tx to the ground station and with a modem rx it could be fed to a computer for analysis or even send the gps information to google maps. What do you need exactly?
Best regards,
Miguel

• ▶ Reply

Permalink Reply by John_Mac on July 30, 2009 at 3:54pm

Thank you Miguel

I would love the system you have!

Right now, however, I need to get the basics running as I described earlier. With that under my belt I would like to take something like this on.

If you have any documentation or pictures, I would like to look over them just to get acquainted with your system.

Best,
John

• ▶ Reply

Permalink Reply by Miguel COLLADOS on July 31, 2009 at 11:38am

Concerning the OSD, here you can see a video and some information: http://alai.h3m.com/~s0350672/catalogo/product_info.php?cPath=43_63... (how do you create a link here?)

And this is the FPV set:
http://alai.h3m.com/~s0350672/catalogo/index.php?cPath=43_67
I think that there is a version of the web page available in english
I have everything installed onto an Easyglider from multiplex, and works ok up to 900m from me.

• ▶ Reply

Permalink Reply by Florin Mingireanu on January 21, 2011 at 2:15am

Hi Bill,

 

Can you indicate where and how the changes should be done in MatrixPilot 2.5.1 in order to raise the sampling rate at 5000 Hz?

In the UAVmanual available online it is also said that the sampling rate should be 5000 and that the servo output rate can be as high as 40 Hz.

 

Thanks!

• ▶ Reply

DeveloperPermalink Reply by William Premerlani on January 23, 2011 at 6:41pm

Hi Florin,

 

I just measured the sampling rate of MatrixPilot 2.5.1, it is 1200 samples per second for each channel.

 

To raise it to 4800 samples per second, change the ADCS field in ADCON3 from 111111 to 001111.

 

You should also change the filtering parameter, FILTERSHIFT, to combine 4 times as many samples as it does now. Change FILTERSHIFT from 3 to 5. If you have a UDB2 and are having trouble with vibration, you might try raising FILTERSHIFT a bit more, maybe to 6 or 7, along with raising the sampling rate from 1200 samples per second to 4800 samples per second.

 

Best regards,

Bill Premerlani

• ▶ Reply

Permalink Reply by John_Mac on August 12, 2009 at 6:33pm

As promised, I have posted a progress report on the application of the UAV DevBoard to a CCPM helicopter. The report describes work Bill and I have done with the Red Board on a TREX 450SE. We are not there yet, but hopefully this gets folks started and capitalize on all the offers of help in developing this.

The new thread is here.

• ▶ Reply

• ‹ Previous
• 1
• …
• 3
• 4
• 5
• 6
• Next ›
• Page