Re: [Gumstix-users] PID control programming strategies

[Dave H. <dhy...@gm...>]

Hi Rodrigo,

> The PID loops algorithm im using make use of information of the past up to 2
> steps backwards (in the actual calculation, it uses info_n, info_n-1 and
> info_n-2 where n is the actual step and info may be the error for example )
> As i may be using about 8 PID loops i was trying to think how to code in an
> efficient manner. My first idea was to define a struct called PID and there
> store values from the different steps, parameters (Kp: poportional gains,
> Ti/Td :Integration time/derivation time, etc), and the actual output of the
> PID loop, in one word, everything.
> Then use functions like PIDinitialize, that would read data from a txt file
> (that way i dont need to recompile if PID parameters are going to change
> during PID tuning)
> and other function like PIDupdate, that performs the next step in the
> calculations.
>
> I would like more the idea of a java class, like storing the parameters and
> previous steps data and perform the next steps calling the function for that
> member (hope i explain well here)

So, I haven't played with PID yet, but I'm familiar with the data
structure stuff.

In java or C++ you create a class and the class has members. In C you
create a structure and pass pointers to the structure around.

An instance of the structure is just like an instance of the class.

typedef struct
{
    int Kp;
    int Kd;
   ....
} PID;

Then you declare some function (i.e. methods) that operate on the structure:

void PID_Initialize( PID *pid );
void PID_Update( PID *pid, int someParam );

In C++ you'd probably do something like:

PID myPid;

myPid.Initialize();

In C you do:

PID myPid;

PID_Initialize( &myPid );

-- 
Dave Hylands
Vancouver, BC, Canada
http://www.DaveHylands.com/