i see bots

In the future, you won't think this is so weird

Category Archives: Reliability

Simple

“Simplicity is the ultimate sophistication. – Leonardo da Vinci (here and here)

I couldn’t pass up the opportunity to use a quote from L. DaVinci as an introduction to this blog post. But it’s really does set the right tone for a topic that’s been creeping up behind me on pad-feet, creaking the floor boards and occasionally breathing a little loudly. It’s always been there, but now it’s time to write about it.

A main theme for home automation should be, I think, to add value – in terms of convenience, security, energy efficiency, etc – while simplifying. Put another way, if you bring a new scenario to bear, don’t do it in such a way as to add complexity. Just add value, not complexity. Hold the complexity! Or make something simpler than it was.

It’s hard to make something simple, or simpler, while adding cool new functionality. Another quote is due:

To paraphrase Einstein (“simply” because I can):

“Make things as simple as possible, but not simpler” – Einstein (here)

Some questions to ponder when considering adding a new scenario:

  • How “natural” does the new scenario seem to the average user?
  • How much “training”, if any, is required?
  • How much of what the user already knows can be used to leverage the new scenario?
  • How robust is the scenario in the face of unexpected user actions or input, perhaps using existing controls or devices? Or other failures (such as power failures, loss of internet connectivity, etc)?
  • How are existing scenarios changed?
  • What additional “workloads” does the new scenario introduced for the user?

For example, some considerations when installing light switches that can be controlled via Home Automation software:

  • Do the switches operate like ordinary switches? Will users just “know” how to use them, because they operate just like other switches around the house?
  • What “value” are you introducing with these fancy new switches… is it for energy conservation? Security? What plan will you put in place to avoid confusing or frustrating users, or, worst-case, leave them tripping in the dark, feeling the walls for the switch for the light that just turned off for some reason?
  • If the HA software is programmed to turn the lights off under certain conditions, how will users react? Can this automation be over-ridden?
  • If the HA software is programmed to turn the lights on under certain conditions, will it also have a plan to turn them off (to conserve power or reduce user workload)?

Consider a (newer) Z-Wave light switch. One of the things I like about Z-Wave is that my controller software (Homeseer) can quickly detect when the user turns the switch on or off (see note 1). This means that I can fire events based on a change in the state of the switch. So I can implement a timer for the switch: if the user turns on the light, I want it to be turned off in, say, 20 minutes, because people in my house don’t know how to turn off lights (or so it seems). So far, so good. But, if someone turns the light on and then off 5 minutes later (that would be me) and THEN turns it back on a minute later, that 20 minute timer needs to be reset. It takes some scripting to make this happen (to clear out any pending “off” events for that switch).

This is an example of using some extra cycles (in the form of additional scripting) to hide the complexity from the user and build in some robustness. It’s also an example of the importance of 2-way communication between devices. It’s hard to create a smart scenario where unpredictable humans are involved if the controls involved aren’t able to communicate bi-directionally. Early Z-Wave devices, and earlier technologies, such as X-10, could not do a good job of keeping the HA software in the loop when the state of the device changed. If the user could turned on a light, the HA software might not know it, or might not know it for several minutes. If you were interested in adding value with these switches, it had to be done carefully to avoid frustrating the user.

Another example: what happens when you “automate” the A/V stack in your living room employing the usual approach of programming a fancy universal remote control (see note 2)? If my household was any indication, you’re introducing a new world of hurt. It seems that the usual approach leaves out the possibility that the user might have the audacity to actually touch the equipment, or perform a step in some order other than what’s been prescribed. This is usually because the typical remote control communicates in 1 direction only: it talks, but does not listen. It has no clue as to whether those commands have been correctly received, or whether it’s own model for the state of the components it thinks it’s controlling is accurate (see note 3). If the hapless user happens to, say, turn on the DVD player by pushing the power button on the player, in order to, say, insert a DVD and then picks up the universal remote to “Play DVD”, confusion results. The hapless universal remote, not knowing that the DVD player is already on will likely send a “power toggle” remote control signal to the DVD player, which will promptly turn if off. The remote is none the wiser. The user, though, is sure that something is screwed up. In an ideal world, the remote and the controlled components would talk to each other. In a slightly-less but still workable world, all component manufacturers would implement discrete remote codes for “power” commands and the like – not toggle commands.

In our house, we’ve moved beyond universal remote controls. At some point, it struck me as to how much of a compromise they represented, in terms of the user experience. Instead of using the remote that came with the component – DVD player, game console, tuner, etc – we were trying to shoe-horn all of the specific functions into a single, large, oddly-shapped remote that also tried to control the components using a 1-way protocol.

It slowly dawned on me that I could take a contrarian approach: if you want to watch a DVD, why not pick up the DVD remote and just use that remote for everything? That remote obviously already has a power button. And, the kicker is, it also has volume up/down buttons – presumably because it can control receivers from the same manufacturer or be pressed into service as a universal remote. So everything one needs to watch a DVD – power, volume, transport, and other DVD-specific functions – are represented as buttons on that one remote. So why not design a DVD scenario around that DVD remote? Similarly, the game console remote had all the buttons one needs to use it as well as a set of volume and mute buttons. Well, this is odd. Let’s go with it.

The approach I took was to reliably and in real-time mirror as software variables in the HA software (virtual devices in Homeseer) the power state of each component. If the DVD player was turned on, I needed the corresponding “DVD Power” virtual device in Homeseer to change instantly to “On”. Ditto the Xbox and any other source components. This was important to get right in order to handle the situation where a pesky human touches something in the A/V stack to, say, switch out a DVD. It took a while to figure out how to do this. For now, I’ll summarize it as follows:

  • DVD player: my particular player, a Sony BluRay player, sports a USB port on the front. When the player is powered on, that port is powered up. I built a simple circuit to sense when +5 volt signal is present at that port and change the state of the “DVD Power” virtual device in Homeseer to “On”. When the +5 volt signal is removed, the virtual device state is set to “Off”.
  • XBox: took a bit of work (and voided the warranty!). The Xbox also has USB ports, but these remain powered on even when the Xbox is “off”. Foiled! So the trick I used for the DVD player wouldn’t work here. I did, however, find a way to tap into the wires leading to the fans, which are fed a varying level of voltage when the console is on (presumably depending on how hot the console is). I built a simple circuit to detect when any positive voltage is present on the fans, and update a “Xbox Power” virtual device in Homeseer accordingly.

Once I had power state variables I could rely on, I was on the way to implementing this “pick up just the remote for the source you want” scenario. If you want to watch a DVD, pick up that remote and hit the power button. The DVD player will turn on, and Homeseer will take note of it: it will run some additional scripts to turn on the Receiver and the Monitor (which entailed making use of the RS-232-based command sets offered by those two components). Enabling the use of the DVD remote’s volume up/down buttons took more work, involving an PC-based IR receiver/transmitter (the USB UIRT).

I’ll post a more detailed explanation of all this in a follow-up post. But the bottom line is that it’s all working now. The existing controls – the remotes, the buttons on the components themselves – still work as expected, and in fact complement the new scenario. The universal remote control is packed away in the closet, beeping now and then as its battery fades.

A final example involves a scenario inspired by a contractor who was working in our bathroom. He kept saying, “Get it out! Get it all out!!”, when we talked about sizing the exhaust fan. He really (really) believed in the importance of clearly the bathroom quickly of steam from the shower. It not, you run the risk of mildew or rot and the resulting structural problems. Always run the fan while showering, and then for 10-15 minutes afterwards.

In my house, however, few other occupants really took this message to heart. If they remembered to turn the fan on, it was typically after the shower was over and the walls were already dripping with condensation. And no one remembered to turn it off after the required 15 minutes, thus triggering complaints from me about noise and wasted electricity.

I moved the fan to a Z-Wave controlled circuit, and scripted it so that it would turn off automatically after 15 minutes. That’s a good start, but doesn’t solve the problem of getting people to turn the fan on during the shower, not after it’s over and the place is already fogged up.

What would the average user expect when asked to describe a “automatic shower fan”? I’d say this: the fan should turn on automatically when someone starts the shower, and not turn off until 15 minutes after the shower ends. That’s a great goal statement. But, as you can imagine, from an implementation perspective, it seemed like a downright gnarly problem to solve. But it was solved, albeit with some extra hardware (Ha7NET hub), 1-Wire devices, and more complex-than-usual scripting). The “simplest solution” has been working well for a couple of years now. More on that later.

In closing: I like ThingM‘s motto: “Smart devices make things simple”. I think that’s a good criteria to evaluate when deciding whether your ‘bots and automation plans are actually adding value.

Notes:

  1. If I’ve set it up correctly. Due to various issues with various versions of Homeseer, it’s not a given that this is always the case. More on this later.
  2. I’ve owned two Logitech Harmony Universal Remotes, but, alas, I can’t recommend them. Both have ended up being disappointed for various reasons… the programming experience is awful, but it pales to the issues caused by the poor product build quality, hardware flakiness, and support. The most recent model I owned is the Harmony 890, which can send signals via IR or RF. If you’re considering one of these, please take a look at the reviews on Amazon first. On paper, it showed promise…
  3. Too often, component manufacturers take the lazy approach to the power button on the remote, by implementing it was a toggle. Pushing the “power” button on the average remote often sends a single command which the component interprets as this: “if you’re on, turn off; if you’re off, turn on”. This means that the average universal remote must remember if a component is on or off in order to implement a scenario like this: “the user is currently watching a DVD but would now like to play a game on the game console… so, turn off the DVD player and turn on the game console”. If the remote thinks that the game console is already on, it will skip sending a “toggle power” remote code. If, on the other hand, the game console has discrete “power on” and “power off” codes, the remote can confidently send “power on”, which do nothing if the game console is already on. I think that “discrete codes” are one of the marks of a component/device that’s designed to be system-integrated.

‘Bots in the Basement

It’s one thing to write about this stuff, but it’s another thing entirely to get hands-on with it. Since I’m after some credibility here :), this post is about some of my own connected experiences, starting in the Home Automation space, in which I started dabbling around 2001.

I try to take a “scenarios-oriented” approach when messing with the house, given that its other occupants may not have a high degree of tolerance for things that don’t work as expected. It’s a goal that any new scenario adds value (in terms of safety, security, energy conservation, comfort, convenience, etc) in a seamless and reliable manner. Sure, sure, sounds good, right?

Here’s a list of scenarios implemented to date:

Lighting

  • Turn off lights if rooms aren’t occupied, or after a certain amount of time
  • If security system is armed, randomly cycle lights to simulate occupancy

Fans

  • Turn off bathroom fans and Kitchen exhaust fan after 10 minutes
  • Turn on bathroom fans if someone is taking a shower, run for 10 minutes after shower ends (this one took a while to pull off, but now it seems perfectly commonplace)
  • Turn on Kitchen fan if high temperature  detected over the stovetop
  • Periodically, run fan in the cats’ litter box area

HVAC

  • Set temperature back in all zones when no one is home (that is, when the security system is armed) and return to normal schedule  when someone comes home (security system is disarmed)
  • Periodically, set correct time on thermostats
  • Periodically, run the hot water recirculation pump

Security

  • Send email notice when security system is armed, disarmed, or when there’s an actually alarm
  • Notify if garage door is left open, provide option for remotely opening/closing it.
  • When someone comes home (and disarms the security system), turn on certain lights for convenience
  • When security system is armed, randomly turn on/off certain lights at night to simulate occupancy

Irrigation

  • Run sprinkers on regular schedule (depending recent rainfall, as detected by local weather station)

A/V

  • In main zone (living room): If DVD is turned on/off, also turn on/off receiver and monitor, and set receiver source accordingly, for main and 2nd zones. Ditto XBox
  • In main zone (living room): If volume up/down is used on any remote, adjust receiver volume
  • In Kitchen zone: if kitchen radio is turned on/off, also turn on receiver zone #2 and set source. Monitor volume keypad for key presses, and adjust zone volume accordingly.

Reminders, Warnings, and Notices

  • Send email note reminding to take out Trash and Compost bins. Scrape city’s web site to also determine if it’s a Recycling day
  • Send nightly reminder note to close apps for better backups
  • If the security system is armed at 10pm at night, send an summary note summarizing the status of the automation system
  • If temperature in computer rack  or over kitchen stove top gets too high, warn
  • Send note if temperate has dropped to near-freezing

These scenarios are implemented via these components:

Homeseer Home Automation server (http://homeseer.com): software installed on a small headless box running 7×24 in the basement. Presently, this represents the main controller and (browser-based) user interface for all home automation scenarios.

6Bit TCP/IP-controlled relay and input board (http://www.6bit.com – appears defunct at this time): I can telnet into this board and command any of its 6 relays to open/close or sense when any of 12 inputs go low. Can also download simple macros, such as: “close relay 1 and open in 10 seconds”. This board is used to: sense when the garage door is open (I mounted a switch on the door opener, a bit of a story in itself), and to sense when certain devices turn on (such as the DVD player, Xbox, or kitchen radio) or when buttons are pushed (such as the volume up/down buttons on the kitchen keypad), or when the security system signals an “armed” / “unarmed”, “normal” or “alarm”, or “motion in zone x”. The relays on the board are used to: close the garage door and energize any of the three sprinkler valves. Homeseer and the 6Bit box communicate with each other via a virtual COM port. The 6bit box also offers an HTTP service, so you can use a browser to configure it, etc. It uses a Lantronix XPort device (http://www.lantronix.com/device-networking/embedded-device-servers/?tab=0) to connect the board/sensor logic to the LAN. It’s too bad that this company appears to be defunct/zombied, as the board is well-designed and constructed. My guess is that if I had to replace it, I’d go the Arduino route.

Ha7NET 1-Wire Ethernet Host Adapter (http://www.embeddeddatasystems.com/HA7Net–Ethernet-1-Wire-Host-Adapter_p_22.html): You plug this small box into your LAN. You also plug into it your “1-Wire” networks. “1-Wire” really is “3-Wires”, but that’s OK, it’s only Marketing. 1-Wire devices are small, cheap, nominally intelligent devices. I use the DS18S20 temperature sensor, which looks like a small transistor and costs < $5. You can plop any number of these devices onto the same run of 3 wires: Signal, Ground, Power. Each device has its own unique ID, and implements a basic protocol for manipulating the 1-Wire bus in order to communicate with the host adapter. I have three little networks of these devices in my house (due to topology requires of the 1-Wire bus), which all terminate at the host adapter. On the Homeseer controller application, there’s a plug-in (from Ultra1Wire) that knows how to find and communicate with the Ha7NET box to get regular (every couple of minutes) updates from the sensors. The Ha7NET box also offers an HTTP service that you can hit with a browser to configure the device, get an inventory of devices on the 1-Wire net, etc. Prior to the Ha7NET board, I used a Midon Design “Temp 08” board (http://midondesign.com/TEMP08/TEMP08.html), that connected to the Homeseer app via serial port. But I wanted to move away from serial-based boards…

Davis Scientific Weather Station (http://www.davisnet.com/weather/products/vantagepro.asp): with a TCP/IP-enabled console. The console communicates with the Davis Scientific “http://WeatherLink.com” web site on a regular basis, sending up weather stats like temperature, wind, etc. The site also sends that info over to any of several popular weather sites such as http://weatherunderground.com. I wrote Homeseer scripts that regularly pull down relevant stats – such as recent rainfall – to guide the sprinkler schedule. So, it’s a bit roundabout  – weather bytes travel from the weather station on my roof, to the console in our kitchen, and there to the WeatherLink site, and then over to WeatherUnderground, and then back down to the Homeseer app running on a PC in my basement. Presumably they’re out of breath when they arrive. An interesting bit of system engineering, but it’s not always reliable (more on that later).

Z-Wave light switches and modules (http://www.z-wave.com/modules/ZwaveStart): Z-Wave is a specification / industry consortium for RF-based switches and controllers that enable remote control and sensing in the house. I have about 10 Z-Wave light switches installed throughout the house, and a Homeseer-branded Z-Wave controller connected to the Homeseer PC via serial port. As a result, Homeseer knows when someone has turned on a light in, say, the family room, and can be scripted to automatically turn if off after, say, 20 minutes. The Z-Wave light switches look mostly like regular Decora-style light switches, with some quirks and other characteristics. Z-Wave devices communicate via an RF mesh… each device spends some time getting to know its neighbors, forming primary and secondary paths to them based on signal strength. When a device wants to communicate to the controller (connected to the PC), it asks its neighbors to route the message along, via the mesh. The advantage in this approach is that the mesh can dynamically deal with sudden obstructions that might block a primary path, such as if a refrigerator door (a large metal object) is opened, or a car pulls into the garage, or permanent obstructions; in my house, one or two of the Z-Wave devices have a clear (RF) view to the controller, as there are all sorts of heating ducts in the way (ever see the movie “Brazil“??). So the mesh concept comes in handy, but it does have its downsides, in terms of complexity, reliability, and ability to debug, which I’ll relate in a later post. But it’s a heck of a lot better than “X10” (http://www.x10.com), an earlier approach (based on sending signals over the AC powerline) to control and sensing in the home. (Some of you may remember how, in the early days of the web and pop-up ads, it seemed that every other ad was for X10…)

Proliphix web-enabled thermostats (http://proliphix.com): Proliphix offers a line of high-quality thermostats that just happen to sport web interfaces and an API. So, you can hit the home page for one of these thermostats and see the current temperature in that room, the current heat/cool setpoints, the setback schedule, and so on. We have three zones in our little house (one for each of the basement, first floor, and second floor), and there’s a plug-in for Homeseer that surfaces key metrics and commands, so that it’s easy to, say, write a script that sets the heat setpoint back if no one’s in the house.

Security System: I’ve installed a mid-level security system, which, until recently, really didn’t want to play well with the other systems. I managed to arrange its typically archaic programming of zones and sensors so that it closes relays for certain conditions (system armed/disarmed, alarm, activity in certain zones), which are connected to the 6bit board’s input sensors, which Homeseer can track. The end-result is that Homeseer knows about certain security system events, and I can script scenarios such as this: if the security system is armed, set back the Proliphix thermostats to save on heating costs.

USB-UIRT IR receiver/transmitter PC interface (http://www.usbuirt.com): this is a small box that connects via USB to the PC running the Homeseer control application, and is apparently manufactured by a guy in a garage :). Homeseer can sense (via a plug-in) when certain IR signals have been received (after training) or transmit IR signals (after training). The result is that I can write scripts like this: “if the volume up button is pressed on the Xbox remote control, then send a signal to the Denon receiver to increment the volume”.

Other elements

I have a Denon receiver with an RS-232 port on the back, which enables communication with a PC or controller. The receiver is connected via serial cable (straight-through! straight-through!!) to the PC running Homeseer. I’ve written scripts for Homeseer which can read responses from, and write commands to, the receiver. So, for instance, someone changes the volume on the receiver – either directly, by turning the knob, or via a remote – the receiver dutifully outputs some characters through the serial port, and my script running on Homeseer can track the changes (updating Homeseer variables which represent the volume level). Similarly, I can, via script, programmatically change the volume on the receiver. Or, I can script Homeseer so that if the DVD player is turned on, the receiver is turned on and its input source set to “DVD”. There’s a similar arrangement with the monitor in the living room. I’ll post more details on how the A/V system is scripted, as I think the resulting level of simplicity it affords to the end-user (no more confusion about remotes, and no more need for universal remotes) is worth sharing.

Some additional, quick notes

  • I’ve tried to move away from serial-based connections (RS-232, USB, etc) and towards TCP/IP, since the resulting simplicity and flexibility (in topology, distance, etc) is worthwhile. The 6bit relay/sensor board, Ha7NET 1-Wire host adapter, and the Proliphix thermostats listed above all communicate via Ethernet. However, I found it much easier to program, and reliable, to stick with the antique RS-232 interface for integrating with the A/V components (receiver, monitor, IR controller). More on this later.
  • I have this hang-up/obsession around the importance of instant and relevant feedback when dealing with the humans in the house (at least, when it comes to the home automation :). If someone presses a button and nothing appears to happen quickly – even if, under the covers, millions of compute cycles are cycled, thousands and thousands of disk seeks are commanded, and umpteen packets are zipped around LAN and WAN, all in service of that human’s simple action – that stinks. The human is left wondering… did I do something wrong? Or did this stupid system fail again? More on this later.
  • Another hang-up/obsession: I don’t have (much) patience for systems that fail, especially those that fail silently. I’ll go out of my way – in the form of extra  coding, hardware, wiring, etc – to ensure that systems are built to be reliable, and that if they do fail, they don’t fail silently. More on this later.

All together, I think this loose constellation of systems and components, held together with lots of Homeseer scripting, does add value in a mostly seamless or even invisible way. There’s lots of room for improvement (for instance, extended power failures can be tough to recover from…). And if viewed from the perspective of, “Is this ready for the mainstream??”, the answer is a clear “No!”. More on this later. 🙂