The document describes an air quality monitoring system using Internet of Things devices. It uses sensors to measure toxic gases, temperature, and humidity, and sends the data via LoRa wireless protocol to The Things Network and then to an InfluxDB database. Tableau software visualizes the timestamped sensor measurements to allow for trend analysis of air quality over time.
2. Goals
• Cost-effective sensors
• Accurate Measurements
• Trend Visualization and Analysis
• Power Saving hardware
• Scalable architecture
• Ready for Smart Grid applications
3. Toxic gas sensor
• P-Nucleo-IKA02A17
for periodic gas
measurements with
temperature
compensation
• Suitable for outdoor
and indoor
environments
4. STM32 LoRa
• ARM Cortex-M0+ core
coupled with a SX176
transceiver
• Enables the board to
use LoRa long-range
protocol
• Used to send the
measurements in the
cloud for further
analysis.
5. STEVAL-MKI141V2 with HTS221
• Relative humidity and
temperature sensor
• Provides further
information about
the air quality.
6. uTensor
• Extremely lightweight ML interface
framework built on Mbed and Tensorflow
• Used to execute an onboard classification
of the measured data.
7. TheThingsNetwork
• Open Network infrastructures to
develop scalable IoT applications.
• Used to manage data flow from the
devices to the cloud platform
8. InfluxDB
• Open source Real-Time Time Series
database used to aggregate data from
several devices
• Data is identified as a timestamp-value
pair
9. Tableau
• Interactive Data Visualization software
• Cross-platform, including a web
interface
• Provides tools to visualize and gain
insight by interacting directly with the
data
10. Workflow
• Measurements collection with the
described sensors
• Data filtering with uTensor
• Transmission of the filtered values over
LoRa and TheThingsNetwork to InfluxDB
• Extraction and visualization of the data
with Tableau
Editor's Notes
The goal of our proof of concept is the creation of an air quality sensor which can give accurate measurements and requires little intervention.
We use inexpensive hardware to make our product deployable with little effort.
Thanks to wireless communication and the hardware cost-effectiveness we can easily place sensors in different places, creating a scalable distributed architecture which provides data to a cloud platform and can be used in smart grid applications, giving information about the environment state.
The gas sensor (Figaro TGS5141) can detect four types of carbon footprints and uses a temperature sensor (STLM20) for temperature compensation, the low-power design of the sensors can guarantee more than 10 years of life time.
This sensor is used to
Humidity accuracy is between +/- 3.5% and 80%, temperature accuracy +/- 0.5 Celsius degrees
Operates between –40 C and 120 C, humidity range from 0% to 100%. The sensor is mounted on a MEMS (Micro Electro Mechanical System)
Given a trained model in Tensorflow it gives a C++ header and source file to execute classification on directly on the board
We use existing network infrastructure (Gateway and applications) to get the data from the board via an IoT-friendly protocol and send them via HTTP to the database