Comment by KM6VV on July 27, 2012 at 10:08am

Hi Chris,

So far I'm running the APM2 "out of the box"; I thought I'd reload it using MAVLink 1.0 after I played with it a little.  So I *think* the APM2 is running ArduPilot, not ArduRover yet.  I had expected to need to download "rover" (APMrover213b). 

I forgot to see what the board signed on with.  Maybe it's just a "test suite" with the 9 or so short tests?  (the program on it asks for a 1 to 9), almost sounds like a Manufacture test suite...  That's starting to make sense.  If that's so, where is the source code for this test suite, I'd like to down load it for later reference?

Alan

3D Robotics Comment by Chris Anderson on July 27, 2012 at 10:13am

Alan, APM doesn't have any software on it all "out of the box" (aside from diagnostics stuff from the factory). You have to load your choice of firmware from the MP before it will do anything. Please follow the manual and everything will be easier :-)

Comment by KM6VV on July 27, 2012 at 10:31am

Chris,

OK, that's what I just realized. It's a manufacturing test suite.  Should MAVLink connect then?  Doesn't it need to connect in order to download?  (OK, instructions say don't connect...)

 

I'd like to save a copy of the "diagnostics stuff" before I do the download.  Is the source code or a bin available? 

OK back to MPInstallation.  Plane, Quad or Heli?  Will we get a pix/mode here sometime?

Thanks for your help!

Alan

3D Robotics Comment by Chris Anderson on July 27, 2012 at 10:42am

I think you're making this harder than you need to. If you want to run a Rover, follow the instructions in the Rover manual. If you just want to test the board via the MP, it doesn't matter which of the firmware types you use. So pick one, plane say, load it and then you can connect via MAVlink. Please follow a manual of some sort and everything will be much easier!

Comment by KM6VV on July 27, 2012 at 10:47am

Yes, I'm "in the manual"

http://code.google.com/p/ardupilot-mega/wiki/MPInstallation

But before I load the Rover and erase the diagnostics, I'd still like the code for the diagnostics, if that's not a problem.  (No surprise, I'm a test engineer!)

Alan

3D Robotics Comment by Chris Anderson on July 27, 2012 at 11:18am

We don't release the factory diagnostic code. It's designed to work with our in-house diagnostic software and won't make any sense outside that. There are dozens of diagnostic tests built into the public ArduCopter/Plane/Rover code (see the CLI Test menu).  

Comment by KM6VV on July 27, 2012 at 11:36am

That's a shame, it's just what I need right now to evaluate the new LEA-6 GPS, and the LOCOSYS GPS I'd hoped to use. 

As I've mentioned before (and didn't really get an answer) I need to be able to walk the "field" with the same GPS in order to save waypoints for the Robomagellan competition.  Rules do not allow one to simply navigate the field with the 'bot beforehand!

The simple exercise of logging the diagnostic output to a file, and cut/pasting my position last night into a map program was immediately useful to me (made a great deal of sense)!

Alan

Comment by Vernon Barry on July 27, 2012 at 12:22pm

Just my 2cents since I too was evaluating the new 6LEA VS the MTK on APM2.0. What's bothered me is that if I take the drone and set it on the ground, no motors and no movement, connect via my 3DR radios to mission planner, it draws in blue the location and movements it "sees" from the GPS. I understand the more satelites, the better the fix, so at first with 3, it may be up to a block away from my actual location. What bothers me is that even with 5 or more, it's walking all over even though it never moved. It looks like a kid drew with crayon on the screen. Obviously if this is where loiter "thinks" it would be in flight then it's not acceptable. The 6 LEA got lock much faster and had 5 satelites right away after the cold start (brand new module from the mail). I actually expected this since the saw filter, remote mounting, and larger antenna help. I still had some movement (up to 2 houses away) sitting still so I chock that up to that's how accurate GPS signal is. Interestly, I tried the MTK now mounted on the external carrier just like the 6LEA and when it got 5 sats, it began to be really steady keeping the sensed location within my back yard.

My general thoughts on the whole thing is that neither GPS is reliable enough to let me set 4 waypoints down my street (I live in a neighborhood around a pond with 4 normal 2 lane streats forming a "box") and have it navigate around the block without hitting trees or houses on either side, let alone loiter above a street corner.

Put it like this, I have pretty wide streets for a neighborhood and a wider back yard than that and it cannot keep the gps "sensed" position in the back yard. If I was wanting a rover or aircraft to go down the street and not hit the curb or trees and houses along the road, it's not happening on current code and the GPS alone (I'm saying the ardcopter code).  I cannot explain why the MTK suddenly seemed better on the second test as I think it was a fluke. The 6 LEA is better, but in general all GPS seems to show drift wider than the typical street.

I also recently watched the DARPA challenge and some of the Sparkfun videos and all the vehicles using GPS as a primary location and waypoints seemed to show the roaming nature of the sensed position (random wander for no good reason).  Maybe the rover code has imporved GPS or maybe others get better results, but right now, today for me, I do not see it possible to navigate down my street on GPS waypoints alone and expect to stay even within the rather wide street.

This leads me to think we must add some "smart" software filtering and maybe more than one GPS to improve accuracy. Maybe this can be done in the module's firmware, but more likely, it must be done in the APM code.

 

Comment by KM6VV on July 27, 2012 at 1:02pm

I probably just got a lucky fix last night up stairs in my office.  Nailed my location in the house! 

A later test down stairs didn't get a fix, and even a trip out onto the deck (under big trees failed).  Oh well!

the 6000 gyro/accelerometer chip can do some sensor fusion, and I understand we now have permission to use some of the aspects of that.  The fusion of inertial, compass and GPS fixes will provide better location data.  We need to extrapolate (dead reckoning) when the GPS is lost.  One of the new LEA-6 chips for automotive use can actually do the dead reckoning internally. 

Funny, I'd expect the "wandering" (other then loosing satellites) to be the same for different GPS chips (assuming it's atmospheric, etc, and not implementation.

I've studied the SparkFun videos and blogs as well.  Quite interesting.

For Robomagellan, getting inside the Waypoint radius will prompt a behavior shift to a vision "blob" tracking mode.  Or LiDAR/SONAR.

Alan 

Comment by u4eake on August 2, 2012 at 2:39am

This probably ain't the right place to put this, but since here seems to be the most activity concerning ArduRover :

I think I found a small bug concerning THROTTLE_SLEW_LIMIT.

In APM_Config_Rover.h you can define THROTTLE_SLEW_LIMIT to get smooth startup throttling, but a little further into the file there is another #define THROTTLE_SLEW_LIMIT 0, which I believe disables the slew limit again.

Haven't tested it yet, maybe this afternoon.

Regards,

U4

Comment

• ‹ Previous
• 1
• …
• 53
• 54
• 55
• 56
• 57
• …
• 78
• Next ›
• Page