Re: Controllerless distributed automation

["David Paterson (Pixi)" <david.paterson@xxxxxxx>]

On Wed, 14 Feb 2001 23:57:06 -0000, you wrote:

>That's a fair point, but it does serve to illustrate part of the
problem,=
=20
>who (or what) decides which neighbours are relevent to each other? - if
a=
=20
>room thermostat does not "know" that the actions of a light
switch in the=
=20
>same room are relevant to it's operation, then might not the heating
stay =
on=20
>way beyond the time the room has ceased to be occupied?

Yep, it's one of the basic problems that I need to solve.  I can either
predefine the "relevance relationships" of device types, but that
might
lead to limitations in what the system can do, or I can design the
system to figure it out for itself, which will probably be impossible to
do effectively.  These initial design choices are going to be the
hardest and most important - and the hardest.

>Actually,=20
>occupancy detection is a really good one to test out any system on, as
thi=
s=20
>has been discussed at >great< length before, and the groups'
collective=20
>expertise has yet to come up with a failsafe system.

I agree that occupancy sensing is a key part of any effective system,
and is one of those problems just waiting for a good, cheap solution.
To me it seems fundamental for ensuring the system is delivering the
right services in the right places at the right times.

I've been interested for a while in the Bik RF systems, but they're no
longer being produced commercially (AFAIK) so that's maybe not a real
option now.  I did hear he'd put the designs into the public domain for
hobby or non- profit use, so I might have a look at them and se if
there's any potential in them.

>But, my point is,=20
>that unless the room thermostat knows the difference between a
thermostat=
=20
>and a light switch, how is it to know that it does not care about
that=20
>difference? - with a distributed intelligence system, you have to
think=20
>about these things. - any component >must< have a knowledge of
it's=20
>environment in order to interact successfully within that
environment...

That's true, but the question I'm trying to resolve is how much
knowledge does it need, and how can I represent it effectively.  I need
to reduce all the different forms of information down to their lowest or
simplest forms, and then build a knowledge base from them.  All a bit
vague at the moment, but discussing it makes me think things through a
bit more carefully.

> (snip)
>
>OK, you got me, - you're right, I am broadly in favour of a
distributed=20
>intelligence model, but only to a certain point. - I don't yet see
the=20
>technology available to implement such a model (at reasonable cost).
You'r=
e=20
>example (frivolous as it may have been ), of a light switch with 1Ghz
CPU=
=20
>and 128MB RAM, will no doubt be with us one day in the
not-too-distant=20
>future, but (and this is the important bit), >NOT< today, and
>NOT< tomorr=
ow=20
>either. (much to my chagrin) (and undoubtedly yours too?)

Well, it wasn't entirely frivolous.  Having watched Moore's law in
action for the past 25 or so years, I can confidently predict that
within 5 years time I'll be able to buy a microcontroller running at
about 200 MHz, with at least 256K of memory, and probably for around =A320
or less.  The 1 gig chip won't take much longer after that.

Of course you'll get the "why would you need all that power?"
perennial
question, but there are always newer applications, smarter ideas and
clever ways to do things which just soak up every drop of processing
resource you have.  You can _never_ have too much power :-)

> (snip)
>
>Keep up the good work, - 'cos if anyone comes up with a solution, it
will =
be=20
>a goldmine!!

Well, I'm not sure I'll succeed, but I'll give it a try.  I've got a
fair bit of relevant (and irrelevant) experience from various jobs doing
things like embedded systems, production line automation, distributed
interactive simulation, and video game development, and there are a lot
of ideas and techniques I've picked up over the years which might be of
some use.

At least I should be able to design a really flashy user interface for
it if nothing else :-)

DP

--=20
David_Paterson =3D david.paterson@xxxxxxx =A6 david.paterson@vissci.=
com;

R & D programmer        There are three kinds of people in the world -
Visual Sciences Ltd.      those who can count, and those who can't...