Fighting PID

[Geek 2.0]

I'm actually in the process of doing that, and just found a strange problem...

I found this in the PIDController class in WPILibJ:

if (((m_totalError + m_error) * m_I < m_maximumOutput)
                        && ((m_totalError + m_error) * m_I > m_minimumOutput))
{
       m_totalError += m_error;
}

m_result = (m_P * m_error + m_I * m_totalError + m_D * (m_error - m_prevError));
m_prevError = m_error;

By the looks of it, it's saying that it won't add the error to the total error if the result would be outside the maximum and minimum output. Does that seem wrong to anyone else? Shouldn't it add the error no matter what, and then clip off the result to fit in the max and min output?

[Patrickwhite]

Originally Posted by Geek 2.0

I'm actually in the process of doing that, and just found a strange problem...

I found this in the PIDController class in WPILibJ:

if (((m_totalError + m_error) * m_I < m_maximumOutput)
                        && ((m_totalError + m_error) * m_I > m_minimumOutput))
{
       m_totalError += m_error;
}

m_result = (m_P * m_error + m_I * m_totalError + m_D * (m_error - m_prevError));
m_prevError = m_error;

By the looks of it, it's saying that it won't add the error to the total error if the result would be outside the maximum and minimum output. Does that seem wrong to anyone else? Shouldn't it add the error no matter what, and then clip off the result to fit in the max and min output?

That is usually how it's done. There are a few things in WPI's implementation that don't make sense, but overall it's fairly useful.

In terms of your original question, even if you used WPI's implementation, your output doesn't have to feed straight into a motor. If you have them as separate outputs, then decide yourself how to feed them into the speed controllers, there shouldn't be any problem.

[kgzak]

We've always written our own PID loops. I have never used the PIDs given to us, I find that if you make you're own you can customize it to what you need. It's not that difficult, I use Labview though.

[Geek 2.0]

I must say, I'm having a bit of trouble understanding what WPILibJ is doing. I'm trying to use it as a reference, but it's contradicting itself. Another example of something that doesn't make much sense:

if (m_continuous)
{
      if (Math.abs(m_error) > (m_maximumInput - m_minimumInput) / 2)
      {
             if (m_error > 0)
             {
                    m_error = m_error - m_maximumInput + m_minimumInput;
             }
             else
             {
                    m_error = m_error 	+ m_maximumInput - m_minimumInput;
              }
       }
}

Continuous means that we don't use maximum and minimum input values. but apparently if we're continuous, we're referencing maximum and minimum inputs values!!! What's going on now??

Maybe I shouldn't use this as a reference...

[Alexander Meyer]

Continuous mode still uses the min and max values, but not as limits--instead, it assumes that the shortest route to the setpoint can be achieved by going past the max back to the min and vice/versa. At least, if the comment on the SetContinuous method is to be believed..

[Geek 2.0]

Wow, just wrote a lot of code (with no testing in between... this'll end poorly).

Thanks, that makes sense. Since I don't have that, I'm not supporting that feature. It all works out now.

[Ether]

Originally Posted by Geek 2.0

By the looks of it, it's saying that it won't add the error to the total error if the result would be outside the maximum and minimum output. Does that seem wrong to anyone else?

Seems right to me.

Shouldn't it add the error no matter what, and then clip off the result to fit in the max and min output?

No. You don't want the integrator error to continue winding up once it's large enough to produce the maximum output.