[racket] A control-theory question

On Wed, Nov 24, 2010 at 4:20 AM, Noel Welsh <noelwelsh at gmail.com> wrote:
> I would adjust the guess by a fraction of the error. This is basically
> a gradient following rule. I.e.:
>
>  Guess(t+1) = Guess(t) + alpha Error
>
> where alpha is some parameter that you set via your simulated annealing

As a followup, the rule of thumb I have picked up is that step 1 in
throwing together a controller should pretty much always be, "Why is
PID not appropriate?"

Typically you can't answer this question, and may stop the design
recipe at step 1. :)

<http://en.wikipedia.org/wiki/PID_controller>

Start with a proportional controller (like Noel's suggestion), if it
doesn't *quite* get you to the set point and that steady state error
is an issue, add in the I and D terms.

Anthony

> On Wed, Nov 24, 2010 at 1:38 AM, Danny Yoo <dyoo at cs.wpi.edu> wrote:
>> Hi everyone,
>>
>> In Moby (my Racket->JS evaluator), I have a parameter that checks the
>> number of steps that the evaluator takes: after a certain threshold,
>> the evaluator temporarily yields control back to the browser, and
>> schedules itself for a restart.  I have a rate at which I want the
>> evaluator to yield --- about five times a second.  However, I don't
>> know how fast my evaluator actually runs per step.
>>
> ...
> _________________________________________________
>  For list-related administrative tasks:
>  http://lists.racket-lang.org/listinfo/users
>