Is HA a solution looking for a problem?

[LarrylLix]

As much as HA is a great hobby and can be very convenient for users, I think the coloured lighting has the issue beat.
 
I don't see any other way to turn on all lights to one of umpteen colours or alternating Christmas theme colours or some other party scheme lighting. One other aspect is that one switch does all 19 lights avoiding the massive time teken to turn on/off all those switches and select colours on each one, if that was even possible. Not a solution to a problem but rather a solution to a new possibility created with new technology that comes with ISY994i / HA.
 
Much of this depends on what you define HA to be. Simple remote control can be done mostly without true HA. There is no automation in it. Same with email notifications. Notifications only on the condition you are not home? Sure.

 

[wkearney99]

It can be hard to defend HA because it seems to take a lot more effort than it should to do things and have them work right.  I'm not even sure why that is.  At a superficial level it doesn't seem like it should be complicated, but somewhere along the way it usually becomes just that.   B)

 
Indeed, it's hard for folks to grasp just how complicated and involved 'fully automated' would really be.  More than just 'involved', more like 'invasive' in that it'd have to know a really terrifying amount of personal details in order to be as truly effective a 'personal assistant' as people desire.  Where you are, what you do, what you've done, who with, what and where you're going to do next... all that and more. 
 
I've often thought of this trove of information as 'quantum data'.  In that, much like the Schrödinger's cat thought experiment.  On one hand you've got an enormous amount of fantastically tempting data in that box.  But merely inferring something from what's in 'that box' would instantly destroy the contents of it.  This is how personal data for behaviors should be treated.  While it'd be terrifically profitable for a business to mine this information, once the public grasped the abuse being perpetrated their model would be at risk of collapse.  Trouble is once the cat's out of the proverbial bag the public may not care enough to stop it.  

So here's the crux of the problem, in order to be truly assistive, such a system would really need to have a on-going relationship with the customer.  That's going to have costs and someone's going to have to pay for them.  Where's the model that makes this work without resorting to pimping the customer out for a gangbang to advertisers?

 

[wkearney99]

This is an interesting read: Do not fold, spindle or mutilate.
 
 

Punch cards became not only a symbol for the computer, but a symbol of alienation. They stood for abstraction, oversimplification and dehumanization. The cards were, it seemed, a twodimensional portrait of people, people abstracted into numbers that machines could use. The cards came to represent a society where it seemed that machines had become more important than people, where people had to change their ways to suit the machines. People weren’t dealing with each other face-to-face, but rather through the medium of the punch card. All of the free-floating anxiety about technology, the information society, “Big Brotherism,” and automation attached itself to punch cards. Examining the metaphorical ways in which punch cards were used lets us understand some of the reaction and resistance to the brave new information world

 
The same sort of thing applies today with use of the web and online services. 

 

[wkearney99]

meanwhile, minority report inches closer: http://www.iflscience.com/technology/now-advertising-billboards-can-read-your-emotions-and-s-just-start

 

[tattema]

Anyway, it's indeed a puzzling question of how to get automation moving forward in interesting ways that don't total revolve around raping the customer's privacy.  While still making a profitable, decent and honest living from it.

 
Hi wkearney, what needs to be done to move automation forward? I can see many things about the current state of HA, namely usability and simplicity, that need to be improved. Is privacy a reason for users not adopting HA in droves?

 

[tattema]

Trev, I agree, aesthetics are very important with the automation either hidden from view or when it is visible it needs to be sleek, matching and integrated with its surrounds. Which is why I have put so much effort into the UI for my widget based HTML5 client and am prototyping a room node which is basically a small 5" touchscreen in the form factor of a light switch. Do share the industrial design for your HA product set (are you marketing it or just for own use?).

Hi Deandob
 
At this time it's just for my own use so but it has a name 'Nimble Cirrus' and I think it could have broader appeal given the simplicity of design. Unfortunately I feel HA is really in need of a solid problem to solve so it might not be commercially viable for the masses. I'm always interested in collaboration so am happy to share (GPL) my designs.
 
I really like your UI widgets and think I'd like to use them in my home. I need a way to render HTML5 on low power low spec MCU's like the LPC1788. Is there a way to do it?
 
Regards
Trevor

 

[deandob]

Hi Trevor,
 
I've been a bit slow to update my client software (been in Europe last week and also a Win 10 rebuild of the server taking longer than it should....). But as part of the rebuild I have setup github correctly so I have source control for several PC's I develop on, so you can access all the code here: https://github.com/deandob
 
There are three repositories:
HAWebClient. This is the HTML5 client with HAClient.html the front page (client framework shell). The files can be dumped on a web server and will work with only the server variable needing to be changed to the server name and have websockets access to the server. You can see the widgets in the widgets directory, and no installation is needed for new widgets, just drop a valid widget file into the widgets directory (use the widgets template file as a starting point).
HAConsole. VB.NET multithreaded code that is the event, message queue, network and automation manager. Uses websockets to communicate and SQLite as the database. Change settings in the settings.ini file
PluginMgr. This is the node.js server that handles all the server plugins (there is also a facility for native dotnet plugins but I find node much easier to develop in). Node manages the web server for the HTML5 client, MQTT services (not fully integrated yet), sockets and REST interfaces and all the plugins. Server side plugins work a little differently to the client, and need a .ini file for plugin definitions for each plugin.
 
See earlier postings for more info on how the system works - although I appreciate you will probably need more info to get it started so just ping me on the forum email. All setup to use Visual Studio Community 2015 (free and very capable IDE).
 
Note - none of this code is 'clean' or production ready, it has a number of hacks, //TODO fixes not done, lacking comprehensive code documentation and portions of it needing rewriting as it has evolved over the years with some aspects better implemented than others. So consider it 'alpha' quality but it does work. If I decide to open source it or even commercialise it I will spend time to clean it up. I do intend to stand up a demo version but need some time to set it up so will be later.
 
Regarding low power MCUs for the server, you would need to run Windows (Intel or Raspberry Pi2) or Linux with Mono. Some of the transforms do quite a bit of work calculating averages and other functions real time on incoming data to have real time 'virtual' devices but this is mostly DB bound rather than CPU, and if you have a couple of transactions a second a small device would be fine. Note that the server is multithreaded and spins out worker threads when needed so ideally you need a dual or more CPU however on my lowly dual core sandy bridge motherboard HAConsole peaks at 3% CPU, node.js for plugin mgr at 5% (3% of the CPU is motion detection from a security camera using opencv in a c++ module). So a quad core Raspberry Pi2 would definitely be suitable.
 
For HTML5 client on a low power MCU, not sure this makes sense as you would need a browser (ie. Windows / Linux) and javascript can be a bit intensive. I see uClinux can be used on the LPC1788 but it isn't full Linux. For a small client I will be running HTML5 on Windows 10 IoT on a Raspberry Pi2 and using the universal app web browser control to host the widgets as part of an application for the small touchscreens I have posted about earlier. Some of the widgets which have automation, video and complex functions like zooming in on a chart will chew CPU and likely won't be responsive on a very small CPU.
 
Do share with me your designs for your form factors, I'm interested to see what you have done.