Developer Comment by Randy on February 13, 2010 at 4:36pm

Nice video. I saw your other one "Above the fog" from 3 months ago as well on vimeo. Very pretty.

Comment by Rana on February 13, 2010 at 7:10pm

Jason, you can well study the code of UDB, there is some height margin defigned, which helps a lot.
@Billu Bhaiya, you can help better.

Admin Comment by Thomas J Coyle III on February 13, 2010 at 7:59pm

Jason,

Like NS Rana said, you probably need to code some hysteresis into the altitude variable so that it takes more than a slight change in altitude to cause the motor to speed up or slow down. It would be sort of like changing your loiter radius until you get the proper response around the loiter coordinates.

Just a thought.

Regards,
TCIII

Admin Comment by John C. on February 13, 2010 at 10:06pm

Wow, very interesting Jason. I had an extremely successful 2.5RC2 flight yesterday, and as happy as I was with it (my first successful throttle integrated flight) I noticed the same notes as you. An oscillation in the altitude hold, and a tendency to figure 8 over home.

I'm having a problem with my KML converter right now, but here's my flight log if you have something that works.

udall_2_12_10.txt


I also have a video that I will post to this thread where you can see the altitude hold behavior.

Excellent work! I can't believe this thing works as well as it does.

JC

Developer Comment by jasonshort on February 13, 2010 at 10:15pm

I'm making some changes that should help on the turns. I'm putting a roll rate limiter to keep my plane from jerking after it reaches a waypoint.

Doug and Ryan are writing a new version of the altitude hold/throttle. I might tweak it for tomorrow though.

Developer Comment by Ryan Beall on February 13, 2010 at 10:27pm

Finally someone believes me with removing the I term! Yeah a better scheme for airspeed/altitude hold would benefit that oscilation. However you kind of have to deal with the gps being noisy unless you add baro. Hmm that makes me think of a filter that you might be able to implement to estimate altitude with out it....I'll see what I can come up with!

Developer Comment by Ryan Beall on February 13, 2010 at 10:28pm

p.s. you can also add a slew rate limiter. That would probably be your best bet to keep this controller for now until a better one replaces it.

Developer Comment by jasonshort on February 13, 2010 at 10:39pm

Yeah, I talked to Doug about the slew rate limiter. I don't exactly know how to write one.
I came up with this:

float calc_nav_roll(float error)
{

        static float err_rate_limiter = 0;

        integrators[I_NAV_ROLL] +=  (error * head_I * deltaMiliSeconds)/1000.0;

        

        // Limit Integrator

        // ----------------

        integrators[I_NAV_ROLL] = constrain(integrators[I_NAV_ROLL], HEAD_I_MIN, HEAD_I_MAX);



        // Sum the errors

        // --------------                       

        error = (head_P * error) + integrators[I_NAV_ROLL];

        

        // rate of change dampener

        if(error > err_rate_limiter){

                err_rate_limiter += 70;

        }else{

                err_rate_limiter -= 70;

        }

        err_rate_limiter = constrain(err_rate_limiter, HEAD_MIN, HEAD_MAX);

        return (err_rate_limiter);

}

70 being .7 degrees
It will slow the change to 35° per second.

I also found I was using some up elevator (5°) to gain altitude. might that be a source of oscillation?

Developer Comment by jasonshort on February 13, 2010 at 10:48pm

Maybe I was over thinking it. Here's an alternate version.

float calc_nav_roll(float error)
{

        static float old_error = 0;

        integrators[I_NAV_ROLL] +=  (error * head_I * deltaMiliSeconds)/1000.0;

        

        // Limit Integrator

        // ----------------

        integrators[I_NAV_ROLL] = constrain(integrators[I_NAV_ROLL], HEAD_I_MIN, HEAD_I_MAX);



        // Sum the errors

        // --------------                       

        error = (head_P * error) + integrators[I_NAV_ROLL];

        error = constrain(error, HEAD_MIN, HEAD_MAX);

        

        old_error = old_error *.9 + error *.1;

        return (old_error);

}

Developer Comment by jasonshort on February 14, 2010 at 12:46am

New code is posted.

Comment

• ‹ Previous
• 1
• 2
• 3
• …
• 5
• Next ›
• Page