Build your own IoT

A short presentation on how I built my own temperature and humidity monitoring system. Presented at WebCamp Slovenia 2017.

Transcript

1. Build your own IoT Because buying things from Amazon is for lazy people. Jernej Virag @jernejv
2. Jernej Virag @jernejv So... why?
3. Jernej Virag @jernejv We had a bit of a humidity issue...
4. Jernej Virag @jernejv ... and off the shelf solutions were a bit expensive ...
5. DIY!
6. Jernej Virag @jernejv Getting the hardware
7. What do we need? • Data-collection device and storage space • Temperature and humidity sensor in each of the rooms • A way to transfer data from sensors to device
8. Collection and storage • One (1) Raspberry Pi (1st gen) with 16GB of SD card • With WiFi adapter (optional) • With screen (optional)
9. Sensors • Turns out sensors aren't that cheap and need: • Microcontroller to plug them into • Power supply • Chasis / box to put them into • Can use a lot of power which causes issues with battery power supply
10. Sensors UGH
11. Jernej Virag @jernejv China saves!
12. Weather stations
13. Weather stations • Sensors already prepackaged • Low power usage (running months on a pair of AAA batteries) • Protected against environment (can be used outdoor even in winter) • Standard 433MHz transcievers • Long range (15m+)
14. Receiving data • Keyword - RTL-SDR (software defined radio) • RTL2832U chip - natively supported in Linux
15. Jernej Virag @jernejv Writing the software
16. Getting sensor data • Collecting data from unknown radio sensors can be annoying - there's no documentation. • BUT.... we're in luck - https://github.com/merbanan/rtl_433
17. rtl_433 console output
18. KISS principle • RPi isn't a fast device - let's make it as lean as possible • Small python script (daemon) running rtl_433 and reading JSON input • Storing data into SQLite database • Small python script serving collected data via JSON • Static HTML/JS user interface served from RPi via nginx
19. Software choice • DietPi Debian distribution running on RPi • Python 3.x • Peewee ORM • Flask web framework • SQLite storage
20. Python daemon • Called tempsd.py • Uses subprocess to run `rtl_433` with JSON as output option • Reads input, parses JSON and saves each reading into SQLite database • Maps sensor ID to room name enum (hardcoded for now =( ) Source on GitHub: https://github.com/izacus/house-temperatures
21. Python server • Called serverd.py • Two JSON endpoints • /current_status - returns current temperatures, time of reading, battery • /graph/<N> - returns hourly grouped data for sensor number N Source on GitHub: https://github.com/izacus/house-temperatures
22. UI • Simple static index.html served from RPi • Dashboard CSS for visual design - https://github.com/hpoydar/dashboard- css • Plotly.js for graphs - https://plot.ly/javascript/ Source on GitHub: https://github.com/izacus/house-temperatures
23. UI Source on GitHub: https://github.com/izacus/house-temperatures
24. Jernej Virag @jernejv What's left?
25. Jernej Virag @jernejv S in IoT is for security
26. What went wrong? • RPi corrupted the SD-card losing all data and breaking the system • Lesson - don't store data on volatile storage, do backups • One of the sensors doesn't support humidity • It'd be great if the solution wouldn't cost more than off-the-shelf stuff
27. // TODO • User interface for the touchscreen • Could be headless Chrome/Firefox running the same index.html and refreshing periodically • Rather light Qt-based interface for better performance • Better data backup system
28. Jernej Virag @jernejv ? 
 
 @jernejv / jernej@virag.si / http://www.virag.si