LoRa sensor grid

Context & objective

My father needed to know when the shop’s fridge or display cases drift out of their temperature range, without running cables through the shop.

Why LoRa: long range on very little power, and a sub-GHz radio that goes through walls better than Wi-Fi without depending on the shop’s network. The Things Network is a free, community-run network server, so no paid platform was needed.

What I built

• Sensor nodes on LilyGO T3 (ESP32 + LoRa) boards and an SX1302 LoRaWAN gateway.
• The gateway’s link to the LoRaWAN network goes through The Things Network.
• Everything after it runs on my own server, in Docker: ChirpStack, an MQTT broker (Mosquitto), Telegraf, InfluxDB for the readings and Grafana for the dashboards.
• Where it stopped: the boards and the whole software chain work, at the prototype stage. I paused before building a finished sensor: developing it needs funding I don’t have, similar products already exist on the market, and the school year was starting, so I chose to focus fully on Epitech.

Challenges

• Nothing on the air. At first my sensor wasn’t reporting through any public gateway, and I didn’t understand why. What unblocked me was talking with the people at the Banyuls-sur-Mer observatory, especially the person in charge of LoRa deployment in the Pyrénées-Orientales, who explained to me how the whole chain works.
• From radio to my data stack: carrying the readings from the radio side into ChirpStack, InfluxDB and Grafana.
• Prototype or product: whether a working prototype is worth turning into a product when it already exists for sale.

What I learned

I expected nothing much, at most a simple radio link. LoRaWAN turned out to be another level.

• It’s a whole network, not a radio. Devices have to join a network server, gateways relay their packets, and if no gateway hears your node, your data simply doesn’t exist.
• Battery life drives the design. How often a node sends, and how much it draws, has to be planned from the start, not fixed afterwards.
• Hardware needs planning. A sensor for a real shop isn’t a dev board on a desk: the enclosure, the power and the placement all have to be designed.
• IoT near food has rules. Monitoring a cold chain means following food safety requirements, not just reading a temperature.
• Knowing when to stop. With no budget and the school year starting, I paused at the prototype to give Epitech my full focus. Now I check those limits before starting a project.