I saw an experimental automated home on TV that was 100% neural. [...] It was a nice system, but SOOOOOOOOO buggy.
[ ..] Since I database Everything, I may at least start some trending. My approach would be [ ..] to make guesses and when it is right, give itself a cookie so to speak and someday if it has enough cookies maybe I would let it try to create some rules on its own. I really think there are too many variables to do much though.
I think until we have accurate People tracking so the system knows exactly who is exactly where, you just can't build much logic automatically.
So Hult, I checked out your website and remember it from a ways back when I was first ramping up my HA research. So where are you at on it these days? What software are you using? Do you have any occupancy logic?
The basic system architecture is unchanged and consists of three 'realms':
I -- 100% Commercial , 3rd-party-serviceable, infrastructure subsystems (Napco Security, Aprilaire/Enerzone/Statnet HVAC, ethernet network, phone system, cable video system, manual lighting). The practical criteria is that if it breaks, one could call any of many different repair companies to fix it. This is much different than being tied to a particular HA/HT installer IMO.
II -- Custom subsystems using commercial software and firmware but custom rule-making (was CyberHouse, now --> HomeSeer +/- CQS; was X-10, now --> INSTEON; was Elk MM443, now ---> Elk M1G; IR Slinke; X-10 RF; etc). Turning these subsytems off should have no serious effect on home functioning, and they should " do no harm" whether enabled or disabled.
III -- Sandbox: Everything else including Work In Progress, both serious and HomeToys . These are in a sandbox for a reason ;-)
(Replacing X-10 with INSTEON has had the practical effect of moving that part of the lighting subsystem from Category II to Category I because INSTEON wall switches/dimmers are dead-ringers for high-end manual equivalents. They could be abandoned as HA component and still function jist fine as manual without bringing attention to themselves. That can't be said for the X10 WS467's they (mostly) replaced ;-)
I was frustrated by hardware limitations in the "people tracking" function. Although CyberHouse has/had some code that was ahead of its time ( eg Motion Vectors), the hard-wired Napco security system turned out to be too slow in reporting through either the RS-232 port or programmed relays to be useful.
One 'fix" would be to use the Elk M1G for a category II occupancy subsystem completely independent of the NAPCO and I've purchased one for that and other applications.
A splendid platform for the Realm III aka Sandbox is National Instruments' www.ni.com both LabView and Visual Studio . I plan to spring for Visual Studio 8.x because I have some non-HA (earth science) applications for it too. NI has robust hardware and software DAQ, real-time and DB presentation, deterministic and stochastic analysis tools with robotics and AI (which NI thinks stands for Automatic Inspection ;-) functions/extensibility that would take lifetimes to duplicate even poorly.
The homebrew ActiveKnob motorized potentiometer lighting system is progressing nicely. It has been fun , and now satisfying, partly owing to the advent/availability of inexpensive and technically manageable tools for microcontroller programming and PCB construction. The motor controller smarts migrated from PC-based to a very small on-motor PCB using AVR ATTiny MCU's.
I'm jist about ready to let loose/post with several more PCBs including THOL ( Temperature Humidity Occupancy and Lighting) sensor boards with redundant, conditioned 1-wire and 0-10vdc outputs, interface boards , potentiometer-mounted controller , battery-backed system power supplies and maybe, some initial AVR BASCOM code.
The PCBs can (mostly) be made and modified with the free version of Eagle PCB software www.cadsoft.de and the AVRs programmed ( mostly) with the free version of BASCOM www.mcselec.com
www.NeuralHome.com .org .net .info .biz