|
Ralph Hempel <rhempel@bmts.com> wrote:
> Mario wrote:
>
> > Paolo Masetti gave everybody a humiliating lesson: his vehicle ran the path
> > in some hundreds less than 10 seconds, at the astonishing speed of 70cm/s
> > (2.3 feet/s). The second best robot took two and a half that time.
> > While running at that speed his robot was playing the Indiana Jones movie
> > theme too :-)
>
> <snip>
>
> > But the most important point, IMO, is that Paolo's robot was the only one
> > that ran a legOs program. I absolutely don't want to say that Paolo won
> > *because* he used legOs, I just want to say his performance was *made
> > possible* by legOs, and in particular by its faster sensor sampling rate.
>
> Alternative firmware is crucial to developing high-performance bots. The
> core sampling rate of the RCX firmware is once every 3 msec, which is pretty
> fast. I'm not sure of the max rate that legOS allows. At that rate, the line
> is sampled every (.70/333) ~= 2mm or .080 in. This is MORE than enough
> resolution for the problem, so the real benefit of alternative firmware is not
> the sampling speed (IMHO) but the *reaction* to the results of the sampling.
>
> The standard firmware must laboriously grind through all of the opcodes
> and watches and evaluate conditions before updating its outputs. Users of
> alternative firmware can optimize this part of the process and make a
> routine that does nothing but adjust steering based on the light sensor, with
> no need to update system timers and other niceties.
I guess you're right this time too :-) Thanks for having clarified the
matter.
> I'll bet that Paolo spent a LOT of time tuning his algorithms to close
> the feedback loop without too much oscillation while still maintaining good
> response, which is one of the classic problems in control systems theory.
Paolo actually confessed he met many problems in fine tuning his bot, and
found a stable configuration only Saturday evening just a few hours before
the race.
Mario
|
|
Message is in Reply To:
4 Messages in This Thread:
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|