|
Posted by D Herring on August 28, 2010, 1:47 am
On 08/13/2010 06:51 AM, Eli wrote:
> Hello,
> I'm looking for the best choice to implement fast PID controllers
> using limited computational resources. The performance of these
> controllers need not be the optimal, but reasonable and off course, it
> should allow easy tuning to stabllity.
> As i understand the I_PARAMETER term is usually created by applying
> the P_PARAMETER to a LP filter. In my case this is too costly in terms
> of computational power, it is much more easy to apply the P_PARAMETER
> into a FIFO buffer and calculate the average value of all the records
> inside this FIFO for each PID Control Tick/
> Is this approach may supply reasonable result?
> The same question about the D_PARAMETER.... if i will just subtract
> the current P_PARAMETER value with an older one taken from the same
> FIFO buffer?
Sorry I've been offline the past couple weeks.
In a PID controller, the control law looks like
u=Kp*P + Kd*D + Ki*I
where
u = control output
Kp = proportional gain
Kd = derivative gain
Ki = integral gain
P = sensor position - desired position
D = time derivative of P
I = time integral of P
As you suggest, a simple approximation of the derivative is
D=(P[t]-P[t-1])/dt. The integral may be calculated by accumulating
I[t]=I[t-1]+P*dt.
where
X[t] = current value of X
X[t-1] = previous value of X
dt = time between the two values
Better values of D may be obtained by using a more sophisticated
derivative filter.
Double-check the above for sanity; it was written rather quickly.
Later,
Daniel
| 
Yahoo! 
Google 
Windows Live 
del.icio.us 
digg 
Netscape 
XML site map
> I'm looking for the best choice to implement fast PID controllers
> using limited computational resources. The performance of these
> controllers need not be the optimal, but reasonable and off course, it
> should allow easy tuning to stabllity.
> As i understand the I_PARAMETER term is usually created by applying
> the P_PARAMETER to a LP filter. In my case this is too costly in terms
> of computational power, it is much more easy to apply the P_PARAMETER
> into a FIFO buffer and calculate the average value of all the records
> inside this FIFO for each PID Control Tick/
> Is this approach may supply reasonable result?
> The same question about the D_PARAMETER.... if i will just subtract
> the current P_PARAMETER value with an older one taken from the same
> FIFO buffer?