RE: Home Automation Laws

["Ian Lowe" <ianlowe@xxxxxxxxxxxxxxxxx>]

Nice theory as far as it goes, however, there=92s a long and ignoble
histor=
y
of that sort of =93intelligent=94 automation.

I like the points made about reducing interaction - for me, I used to do a
stack of things to set up the home cinema for watching a movie: you know
th=
e
usual, turn on the projector, switch on the subwoofer, turn off the other
audio sources in the kitchen, close the curtains, dim the lights...=20

Now, I press a single button - and it all happens.

Thing is, there's nothing intuitive about it, not even close. Rather than
pressing a mystical button, I might as well be waving a wand, whispering a
password etc..

For a visitor, it's obtuse.

Switching the living room into movie mode is just one example of course,
another would be my "night time" routine. It has been refined
over many
attempts, and variously switches off things like the lights in the kitchen,
the computer room lights, the amp and speakers in the kitchen, turns on the
'leccy blankets, turns on the bedroom lamp, and dims it to a nice welcoming
level...

The problem is, of course, that our night time routine isn't as set in
ston=
e
as you think... some nights we don't want to go to bed early, some nights
w=
e
want to go to bed really early, if you catch my drift, and really don't
wan=
t
the whole macro thing to kick off around us...

So you end up with more and more and more requirement to take into account
the sorts of stuff that we humans take for granted, but all involve more
inputs, more sensors, more of everything - all of which have a not
insignificant cost attached. There are a very large number of variables,
an=
d
each variable has a tangible cost connected to it.

It's one of my foibles about the web brick. I reckon that to do occupancy
detection *properly* for my fairly modest 3 bedroom semi, I'd need a couple
of hundred or so digital inputs to a system, not 16, not even 64 -
hundreds=
.

And even then, we'd probably find the need for another sensor, switch, pir
or contact, to help us distinguish ever more precise sets of
circumstances...

It's easy to describe the principle of a learning HA environment. I
sketche=
d
out a system of loadings and weightings on the outputs and inputs using
*very* simple stimuli-response reactions - a simple neural net using one of
the very good code samples out there for this sort of thing, but gave up,
largely because it was obvious that the number of inputs needed to make any
kind of sensible decision was absolutely HUGE...

That's what makes me think that the technical feasibility of this is, imo,
just not there yet. Processing is cheap and getting cheaper all the time,
s=
o
it's getting there. The number of smart Ethernet nodes that have popped up
in recent months is quite thrilling, with ever lower price points.. when we
have Ethernet connected, IP talking nodes with 64 and 128 I/O lines
available for =A320 a throw... then we can have the richness of sensor
information available to start doing really smart stuff.

Phew. That's a long email.

Ian.




** Sponsored by http://www.BERBLE.com **
all the Cool Stuff, in one Place=20