
Over the year's I've tried almost every possible home automation sensor to determin which ones
work well and provde to be reliable.

Uses for sensors can be divided roughly into people motion sensing, people presence sensing, environment sensing (temperature, pressure, solar input, gas and particulates) and indirect sensing.

Indirect sensing is a useful technique for monitoring equipment without making any electrical connection to it. For example you can check that a fridge is working reliably and that the door has not been left open by monitoring the temperature on the motor or cooling coil. If it cycles regularly the fridge is working. If it's above average for a long period then the door is open or there is some other problem with the fridge.

In choosing a sensor for home automation there are many factors you should consider:

1. Reliability - my top criteria because a sensor that fails even occasionally is useless.

2. Accuracy - for sensors that read analog values what is their precision and how stable are they?

3. Latency - how quickly does the sensor fire after motion has happened. Some [PIR sensors](/posts/sensors/pir) wait until
they have seen three pulses before firing.

4. Repeat rate - how quickly the sensor recovers and is willing to fire again. Some battery-powered alarm sensors wait up to three minutes between firing in order to conserve battery power.

5. Noise or False trigger susceptibility - [Microwave-doppler radar sensors](/posts/sensors/microwave-doppler) provide great coverage and are cheap but they fire for no apparent reason. Some humidity sensors are quite random in their output and require severe filtering.

6. Person-discrimination capability - can the sensor tell the difference between a human and a dog or cat?

7. Wiring complexity - the one-wire bus makes it easy to link many sensors with a single very long CAT5 cable.

8. Aesthetics: most PIR sensors are ugly whereas under-floor strain-gauges are invisible.

9. Power consumption - not a factor except for some remote battery operated sensors.

People-detecting sensors
----

In terms of people-detecting sensors I would rank them as follows:

1. Under-floor strain gauges: invisible, very reliable, good cat and dog discrimination powers.
2. Physical contacts on doors and windows - the typical alarm system sensor.
3. Optical-beam with two or more beams: can cover a large gap like a garage entrance or barn aisleway.
4. PIR - cheap and readily available.
5. Bluetooth - the only sensor that can count people (phones) in a room without a camera.
6. Microwave-doppler - cheap but needs logic to clean up false triggers and to reason about which rooms are occupied.
7. Indirect sensing - CO2, temperature, pressure can all aid in detecting the presence of people

Essential sensors
----

In addition to people sensing you need several other key inputs for a complete home automation system, here's my top list for those:

1. Outdoor light sensor. A simple LDR suffices but you can also get multi-spectral sensors.
2. Temperature sensors for living spaces. Essential for any control of HVAC systems.
3. Humidity sensors. These are useful for controlling bathroom fans and preventing condensation.
4. Temperature sensors for equipement. DS18B20 sensors on a one-wire bus are my favorite here given their reliability, accuracy and ease of wiring.
5. Gas sensors. Detecting CO2 levels in a room can help trigger ventilation.
6. Pressure sensors.


An overview of the many sensors I've experimented with for home automation including my favorite under-floor strain gauge, through all the usual PIR, beam and contact sensors to some more esoteric devices like an 8x8 thermal camera.

Sensors

Home Automation Sensors


Even the more reliable sensors will glitch occasionally and some of the less reliable sensors will glitch all the time. Filtering sensor data to ensure that these bad values don't get through is essential. Here are some techniques:

Moving Average
----

This has reasonable results and is very simple to code. You can maintain a circular buffer of values and calculate an average each time you add a new value by averaging all the values in the buffer. In it's simplest for you can keep just one value and do something like:

````c
   current = 0.9 * current + x * 0.1
````

But it's not great. The glitch values still move the average, and the output lags the input.

Median filter
----

Instead of taking the average of the buffer, take the median value. This means you need to copy and sort the buffer but that's fairly easy even in C. This approach also lags but it is great for rejecting the occasional stupid spike.

Kalman filter
----

The Kalman filter is a bit harder to understand but essentially it works a bit like a moving average but it also learns how much of the new value to bring in each time and how much of the old value to maintain. It can also operate on multidimensional data.

Recommendation
----

In addition to rejecting just plain stupid outliers (e.g. 212 degrees for an indoor temperature), I recommend a small median filter to reject outliers followed by a kalman filter for most situations.


Filtering raw data is essential for a reliable home automation system. Here are some of the many ways you can filter sensor data.

Filtering techniques


Traditionally iBeacon technology was created to allow a mobile phone to know where it is in a building. A museum for example might place an iBeacon on each display and have an app that tells you about the exhibit in front of you. This works well but you need to have your phone on and you need an app running for this to be useful. When your phone is locked it still checks periodically for location beacons but the interval is around 15 minutes and is therefore useless for home automation.

Instead, my interest is in using the Bluetooth advertisements sent by cellphones, Apple watches and iBeacons to track where people, dogs and assets are located within the home. I see this data as complementary to all the other sensor data collected and plan to fuse the two in the future.

Phones however do not want to be tracked, they include a technology called "mac address randomization" that causes the phone to skip to a new mac address on a random time interval. iBeacons and many other devices are much easier to track.

My approach has been to install multiple Raspberry Pi devices around the home and barn and then use code to figure out where each phone, watch or beacon is. The code includes hundreds of heuristics to discern the device ID from any information it transits including manufacturer id, manufacturer data, mac address (if public), appearance, name, ...

Stitching mac address randomization back up into a coherent track has been achieved! I can now track a person with a phone as they move through the space from room to room even as their phone skips mac addresses. Only when two phones in the same room switch mac address at the same time does the system lose the ability to track an individual but it retains the ability to know how many people are in each room.

Determining the location from Bluetooth RSSI has also been a challenge. The next version of Bluetooth will include direction finding capability to centimeter accuracy using angle-of-arrival, but I don't have that yet. Mapping RSSI values onto locations is complex because reflections (off metal surfaces), attenuations (walls, furniture, ...) and atompsheric conditions (mostly humidity) can all affect the received signal strength. Further complicating the issue is that fact that not all of the Pis will receive any given advertisement packet and some of the Pis may be offline for maintenance.

Despite all these problems I now have a robust algoritm that can map devices to rooms, and in some cases areas within a room.

Applications
----

I envision many different uses for the Bluetooth tracking technology I have developed including:-

1. Occupancy-based heating control, e.g. lower temperature when just one person present.
2. Lighting controls can now be more certain that someone is still in a room even when they do not move.
3. Asset tracking enables "find my anything" for any Bluetooth tagged tool or dog.

Trials
----

The technology is currently in trials in the Kuwait, Baltimore and Seattle. A huge volume of data is being collected every day to improve the heuristics for device categorization and to tune the location tracking aspects of the system.

Awards
----

The technology has won two awards so far, one from Hackster.io and another global hackathon with a $20,000 first place prize from HCL. You can read more about that on [Geekwire](https://www.geekwire.com/2020/avoid-crowd-seattle-area-inventor-wins-global-hackathon-system-monitors-indoor-spaces/).


Bluetooth sensing for home automation is a great proxy for people counting as it can detect and locate each cellphone in the house. iBeacons attached to tools, cars and pets can provide a 'find my anything' feature too.

Bluetooth Sensing for Home Automation


I currently have a VOC sensor in the garage, it provides some 'interesting' data but nothing useful yet.

![Gas readings](/images/sensors/gas-sensor-graph.png)

I recently acquired a CO2 sensor and plan to add that to the system.

A particulate sensor would be a good addition too for the air handling system.


Gas sensors come in many different flavors including CO2, VOC and particulate sensors.

Gas sensors


The DHT11 and equivalent humidity sensors can be connected to an Arduino or ESP8266 easily to provide humidity data to a home automation system. The local pull-up resistor is critical: reduce it to ensure you get a reliable digital signal from the device. Connecting them over a long distance has proven tricky as there is a lot of noise on the data line. Even with the correct resistors and short runs I find they still provide bad data occasionally (drop or add a bit) so I also feed them into a simple filter: keep the last 5 values, sort them by value and then pick the middle one.

I use humidity sensors in the bathrooms to automatically turn the fan on and the heating element behind the mirrors when the bathroom becomes steamy.


Humidity sensors are great for controlling extractor fans in bathrooms and other damp spaces.

Humidity Sensors (DHT11, DHT22, AM2320)


A simple light-dependent resisor (LDR) connected to an Arduino posts regular updates with the light level outside. More sophisticated light sensors are available including ones that give readings on different wavelengths allowing for compensation from UV light. I've tried them but they didn't seem any more useful than the simple LDR for determining how sunny it is outside.

Placing the light sensor is a bit tricky because you want an unobstructed view of outside where there are no house lights that could affect the reading. A single tree in the path of the sun also creates an interesting dip in the graph but generally not enough to affect the determination 'sunny' or 'dark'.

The home automation system uses the external light level to trigger internal lights to come on according to the [light state machine](/posts/2018-02-2018-02-10-state-machine-for-lightining-control/). Each light has its own percentage value at which the lights should come on and the level at which they should go off.

This works really well in practice. You can be sitting in the kitchen as a storm rolls in outside and the lights come on just as you notice that it's getting dark inside. In the morning the same lights go off as soon as it's bright enough that they aren't needed. Different lights in the same room can have different percentages set to they come on in a staggered fashion as it gets darker.

Overall I would recommend adding a solar sensor as early as you can if you are creating your own home automation system.

Future plans
----

I plan on adding at least one more solar sensor to my set up to measure the light in different compass directions and to provide some fault-tolerance.

Snow warning
----

Recently we had snow on the ground (a rare occurrence in Seattle). It reflected a lot of light up at the solar sensor so it appeared extra sunny outside, however the skylights were covered with snow so those rooms were darker than normal. Even after 15 years I'm still finding edge-case scenarios that the system can't handle perfectly.


Having at least one light sensor is critical for any home automation system that controls lightng. Lights need to be turned on when it's dark not at specific times of day, especially here in Seattle when it can be dark and cloudy at any time of day.

Light sensors for Home Automation


The RCWL-0516 microwave doppler radio board is a very cheap sensor for home automation. It responds to even slight motion anywhere within a considerable range (~10m) and it works through walls and other obstacles. It's not really "doppler" but it does detect changes in the interference between a weak transmitted microwave signal and the returned signal.

The downside to these properties is that it's no longer telling you about occupancy in a single room and it's susceptible to random triggers for no apparent reason. I tried the various hacks to reduce the sensitivity but it still triggers occasionally when it should not.

Given these properties I was forced to come up with a new approach in my home automation system that can handle unreliable sensors like this and sensors that span multiple rooms. My probabilistic model solved both these problems. By assigning a low probability to the radar sensor type and by associating each radar sensor with a percentage coverage in multiple rooms I can incorporate the useful data it provides whilst ignoring the spurious triggers.

In effect the radar sensor can confirm that someone is still in a room and it can detect someone has moved to a room (after multiple triggers each one increasing the likelihood of a move), but it cannot on a single trigger determine that someone has entered a room, nor even which room is being triggered.

The sensors are cheap and worth including in your home automation plans. Some commercial sensors combine technology like this with PIR to provide greater false rejection for both sensors which seems like a smart approach.


Microwave doppler sensors can be found in some alarm sensors but there are also available very cheaply as a separate component. They offer exceptional range but suffer from false triggers requiring a probailistic approach to people sensing.

Microwave Doppler Sensors (RCWL-0516)


Optical beam sensors are reliable and can cover a long-distance such as across a garage or aisle-way. When they include multiple-beams they have good false-trigger rejection. Typically they run on a 12v power supply and connect into an alarm panel.

Initially I had a single-beam optical sensor across our barn aisle but the IR sender seemed to atrract moths at night and they were capable of triggering the beam. A dual-beam sensor (separated about 20cm apart vertically) provides a much more reliable sensor and the only false alarm I get now is from a large crow that sometimes flies through the barn.

They are suitable for indoor, exposed or outdoor applications but are too ugly for most indoor applications.


Optical-beam sensors are reliable and can cover a long-distance such as across a garage or aisle-way. When they include multiple-beams they have good false-trigger rejection.

Optical-beam sensors


PIR sensors are cheap and easy to use but they suffer from slow response times and low repeat rates.

A PIR sensor works by detecting the thermal radiation given off by a warm person as they move across the field of view. Using a fresnel lens they convert the motion into a changing level at the sensor which triggers it.

Often they have a switch inside for 1 or 3 pulses before they trigger. I always move this to 1 pulse so that I get a faster signal and greater sensitivity to movement. The downside to doing that is false triggers but I handle that with the probabilistic treatement I apply to all sensors.

Battery-powered PIR sensors for alarm panels may have repeat rates as low as once every three minutes. This is fine for an alarm panel but somewhat useless for home automation, avoid them if you can. They are however worth adding in places where you otherwise can't add a sensor because there is no power source or wiring.


PIR sensors are cheap and easy to use but they suffer from slow response times and low repeat rates.

PIR Sensors for Home Automation


I've been using Pulsor under-floor strain gauges for many years. Currently they are connected through an alarm panel into the home automation system but I plan on connecting them directly to an HX711 and reading analog values from them.

Strain-gauges vary their resistance as they flex. By gluing them to the underside or side of a joist they can detect someone standing in an area about 2m x 1m above. The exact area covered depends on the direction of the joists and the stiffness of the floor but even through a hardwood floor and 0.75" subfloor they can detect an adult above.

If you need more sensitivity you can put a very shallow saw cut in the beam and glue them across the gap, this concentrates the stress on the sensor.

By wiring them in a wheatstone bridge you can compensate for any temperature sensors. An ideal set up has four sensors connected together but two with 1k ohm resistors works well too and if necessary a single sensor and three 1k ohm resistors will work.

Pulsor sensors are also ideal for situations where a regular PIR sensor cannot work. People use them on boats where the rocking motion and sunlight would trigger many other sensor types. I use them under outdoor decks and indoors on stairs and some room floors. The biggest issue is that you need access under the floor to install them.

Long term plans include using the HX711 (a digital scale chip with an accurate ADC) to read them. Can I detect who is above by their weight? or the shape of the impulse as they walk over the sensor?


Strain-gauges are my top-rated sensor for home automation because they are invisible, reliable and can be tuned to detect people and ignore pets.

Strain-gauges


My favorite temperature sensor is the DS18B20 which connects to a one-wire bus. They are available in a transistor-sized package for board mounting or in a waterproof metal tube. They cost around $2 each and you can string multiple sensors together on a long wire with just three conductors. The signal is digital so there's no issues with random noise on the wire and they have proven to be exceptionally reliable.

I use them to monitor every pipe on the boiler system, in-flow and out-flow temperature; the fridge and freezer operation; the hot tub and some external temperatures.

Even though the DS18B20 is more reliable than the [DHT11/22](humidity) sensors it may still glitch:

![Freezer sensor glitch](/images/sensors/DS18B20Glitch.png)

So you'll want to apply one of the [filtering techniques](filtering-techniques) to the data.


Temperature sensors I've experimented with for home automation.

Temperature sensors for home automation


Coming soon, discussion on the 8x8 grid-eye sensor, an affordable compromise between a camera and privacy concerns.


Experiments with an 8x8 IR camera for privacy-preserving people detection using cameras.

The Grideye 8x8 camera sensor

CCTV cameras are an option for detecting people but within the home there are privacy concerns that need to be addressed.

CCTV Cameras as Home Automation Sensors


I haven't found many uses for these yet, potentially detecting clogged filters through an increase in differential pressure. 

Pressure sensors can detect HVAC system operation and could potentially detect clogged filters.

The Blog of Ian Mercer.

Topic: sensors