Posted by Jordi Muñoz on August 11, 2009 at 4:30pm
Whats new:-Upgrade the navigation algorithms for the REAL ones.-Fixed the airspeed (yeah the problem was the groundstation not the ardupilot itself).If you have an ardustation download the version >= 1.2 please.-I added the RSP (Roll Set Point) variable to know if the NAV is doing the right job.-I added the TEST_SENSORS in line 5-8. This will inject the orientation data comingfrom the IR’s sensors to the servos, so you can test fast and easily if you setup is correct.If the rudder/ailerons and elevators moves contrary to the aircraft orientation means you’re are OK.-I added the CALIBRATE_SERVOS in line 5-7. Will let you see the range of your servos(+-45 degrees). For example in my Hitec HS-81 the full range of +-90 degrees is 600ms and 2400ms,this values must be set in SERVOS MAX and MIN located in lines 3-1 to 3-4 (now are in depended forailerons and elevator), you should see you servo moving perfectly 45 degrees (IF NOT, ADJUST IT!).-Walk Around mod. Line 5-6, If is set to 0 will let you test visually the navigation.Actually will disable the stabilization and will pulse the rudder/ailerons servo directly.You can also use it as ArduPilot V10 (is what actually does to navigate).Ardupilot V10 can be killed now.-FIxed the Header comments (Chris actually did it).Tricky, Tricky!I have found some Tricks:-If you don’t connect the receiver to the ardupilot it will get stuck when the GPS is not fix and during the start up.This thanks to the PulseIn function (not my fault).-Note in this version the PI gains are not calibrated i need to tune them again.-I've change the default servo polarity (plug and play in my setup (= ) Nothing personal really.ArduPilot23_RC1.zip
I previously attached a pic in the forum post about the ground behavior- http://snipurl.com/qbflj I'd be willing to bet that the XY sensor couldn't be more ideally located than where I have it (top of the vert stab). I also have a mockup fuselage with a 2nd complete setup on it, and it does exactly the same thing as the one on the plane, so I think hardware can be ruled out.
The field setup procedure I'm using is straight out of the new manual (which is excellent BTW). The procedure says to hit the reset after getting a lock, but does NOT say anything about holding the plane up level when you do it. It makes perfect sense that at some point we need to teach the system what "level" looks like.
Unfortunately, I just tried holding the mockup high and level in the yard when I hit reset, and it made no difference. I turned around 180 degrees and tried again, just in case the obstacles caused it to be wrong, but it was the same. I also tried pointing it at the ground, and also at the sky, with no change. It just doesn't appear to be setting "level", unless this isn't the reset you were talking about?
I had mentioned on the other thread that I never installed the config tool, since my intention was to test RTL first, and I assumed the config tool wasn't needed for that. Skimming through the manual, I noticed this:
""Roll Trim" and "Elevator Trim" are to set your autopilot to compensate for aircraft tendencies to drift one way or another. This is most useful if your XY sensor isn't mounted quite level. "
Hmmm, maybe I should install the software and try to look at this. I wonder if there could just be a huge trim value in there by default??? I'll try to take a look at that now.
Can you send a picture of your setup? You have have the XY sensor in a poor location.
I assume you're doing the setup procedure right at the field, including the reset when you're holding the plane in the air. None of us are seeing the problem with our setups, so it does some like something particular to yours.
I have a Twinstar, with RC1, including the changes that Chris mentioned. On the ground, it seems to navigate properly when walking around, and the thermal sensors have the correct reactions. The only thing that seems wrong is a significant down elevator movement when engaging the AP. I asked about this on another thread, but no one really had a definitive answer to whether that was normal.
Minutes ago, I took it up for a test flight, and can confirm that it does exactly the same thing when engaged in flight. Basically, it assumes an immediate course towards the core of the earth :-) I tried it twice, with similar results. Once it nosed straight down, it's happy to stay on that course, and you can see the same thing when holding the plane on the ground. The elevator will start reacting upward as you pitch forward, but it doesn't get to neutral until you're pointing nearly straight down.
I have two complete setups, and I believe I've seen this same behavior on both, but I'll be confirming that in a few minutes. It's as if it doesn't know what level is supposed to be. Is there a setting that is used to adjust level, such as if you were forced to install your xy sensor in an unlevel location? If so, maybe mine is wrong.
O.K. I get it. In other words you can use 2.3 with stand alone stabilization.
Seeing that Ardupilot 1 sold over 300 copies if I'm correct, it will be good news to some out there if they can use 2.3 but code will have to be trimmed to fit Ardupilot 1 right?
Comments
I previously attached a pic in the forum post about the ground behavior- http://snipurl.com/qbflj I'd be willing to bet that the XY sensor couldn't be more ideally located than where I have it (top of the vert stab). I also have a mockup fuselage with a 2nd complete setup on it, and it does exactly the same thing as the one on the plane, so I think hardware can be ruled out.
The field setup procedure I'm using is straight out of the new manual (which is excellent BTW). The procedure says to hit the reset after getting a lock, but does NOT say anything about holding the plane up level when you do it. It makes perfect sense that at some point we need to teach the system what "level" looks like.
Unfortunately, I just tried holding the mockup high and level in the yard when I hit reset, and it made no difference. I turned around 180 degrees and tried again, just in case the obstacles caused it to be wrong, but it was the same. I also tried pointing it at the ground, and also at the sky, with no change. It just doesn't appear to be setting "level", unless this isn't the reset you were talking about?
I had mentioned on the other thread that I never installed the config tool, since my intention was to test RTL first, and I assumed the config tool wasn't needed for that. Skimming through the manual, I noticed this:
""Roll Trim" and "Elevator Trim" are to set your autopilot to compensate for aircraft tendencies to drift one way or another. This is most useful if your XY sensor isn't mounted quite level. "
Hmmm, maybe I should install the software and try to look at this. I wonder if there could just be a huge trim value in there by default??? I'll try to take a look at that now.
Thanks,
Rusty
I assume you're doing the setup procedure right at the field, including the reset when you're holding the plane in the air. None of us are seeing the problem with our setups, so it does some like something particular to yours.
I have a Twinstar, with RC1, including the changes that Chris mentioned. On the ground, it seems to navigate properly when walking around, and the thermal sensors have the correct reactions. The only thing that seems wrong is a significant down elevator movement when engaging the AP. I asked about this on another thread, but no one really had a definitive answer to whether that was normal.
Minutes ago, I took it up for a test flight, and can confirm that it does exactly the same thing when engaged in flight. Basically, it assumes an immediate course towards the core of the earth :-) I tried it twice, with similar results. Once it nosed straight down, it's happy to stay on that course, and you can see the same thing when holding the plane on the ground. The elevator will start reacting upward as you pitch forward, but it doesn't get to neutral until you're pointing nearly straight down.
I have two complete setups, and I believe I've seen this same behavior on both, but I'll be confirming that in a few minutes. It's as if it doesn't know what level is supposed to be. Is there a setting that is used to adjust level, such as if you were forced to install your xy sensor in an unlevel location? If so, maybe mine is wrong.
Thanks,
Rusty
--The default GPS in the .h file was accidentally set to uBlox. Change that back to SIRF if you've got a EM406
--There's a typo in the .h file:
" //3-17
#define pitch_P 65 //Pitch Proportional"
That should be .65, not 65!
-Peter
Seeing that Ardupilot 1 sold over 300 copies if I'm correct, it will be good news to some out there if they can use 2.3 but code will have to be trimmed to fit Ardupilot 1 right?
-
1
-
2
-
3
of 3 Next