(Work in progress)

• I have not been thinking of BLE as wireless transport for IoT so far, but Bluetooth 5 has some enhancements for IoT: mesh networking and AoA(Angle of Arrival).
• I saw Bluetooth router products developed by Cassia Networks at a trade show held in Tokyo in April 2019. I got that BLE is not only for 1:1 but also 1:N.
• RasPi (BlueZ) supports BLE, and its Python wrapper, bluepy, is very easy to use.

• RasPi has enough processing power for handling IoT time-series data, so why not just use it as an IoT platform.
• RasPi has four cores, so it can handle data from mulitple network interfaces in paralell.
• If its I/O bottleneck matters, connect SSD to it.

• Dynamic NFC tag (RF+I2C) can provide rich user interface to the user of this router.

• Develop a router that bridges between smart phones and IoT/EdgeAI.
• Support low-power wireless transport: BLE (main), EnOcean, NFC and TWELITE(IEEE802.15.4-based).
• Very easy to install or support ad-hoc deployment.
• Support IoT/AI applications as SPA and as standalone processes (or as Docker containers).
• Support device management for the router and IoT/AI nodes.

                   ============== ROUTER ===============
   Chrome          . . . . . . RasPi . . . . . . . . . .                                      STM32 or PIC16F1
[spa(Vue.js)]<----->[mosquitto]<--->[gateway.py]<--->[Comm. module]<- Low-power wireless - >[IoT node or edge AI]
        ^          :     |                             :             BLE, EnOcean, NFC etc
        |          :     V                             :
        +----------->[api.js(node/express)]--[sqlite3] :
                   :     |                             : 
                   : [IoT/AI applications]             :       
                   . . . . . . . . . . . . . . . . . . .                                   

Single board PCs such as RasPi. In other words, Ubuntu Linux on Four-core Arm Cortex-A53 MPU at 1.2GHz.

        +---------+
        |         |<---[gateway.py]<----[BLE: Microchip RN4020] (American)
    <---|mosquitto|<---[gateway.js]<----[EnOcean: EnOcean USB400J] (German)
        |         |<---[gateway.py]<----[TWELITE/IEEE802.15.4: MONO WIRELESS MONOSTICK] (Japanese)
        +---------+
        
       Note: NFC device is remote-controlled via BLE.

My conclusion in this project is that Microchip RN4020 BLE module (Bluetooth 4.2 BLE) is the most useful wireless module for IoT prototyping. RN4020 is also easy to use.

==> BLE (Microchip RN4020)

==> EnOcean (EnOcean GmbH)

==> TWELITE (MONO WIRELESS)

MQTT topic: "sensor/<source device name>"
MQTT message format: "<source device name>,<timestamp(:.3f)>,<data0>,<data1>,..."

Time stamp format: .

Note: when it comes to wireless IoT, my experiences in my past IoT projects proves that MQTT messaging should use a CSV format (or binary) rather than JSON that adds a lot of overhead.

Request:

MQTT topic: "<destination device name>/rx"
MQTT message format: <command>

Response:

MQTT topic: "<destination device name>/tx"
MQTT message format: <response>

• In this project, I use SQLite, because I want to run everything on RasPi.
• I use node.js as a web framework to implement RESTful API server.

==> RESTful API server with sqlite3

The router provides HTML5-based applications:

• Sensor data visualization

The router reads/writes NDEF data on NFC tag via BLE and I2C:

• Dynamic NFC tag with BLE and I2C
• HTML5 GUI to write NDEF data to dynamic NFC tag via BLE and I2C

• Low-end edge AI for acoustic scene classification
• Dynamic NFC tag with RN4020(BLE module)

• PTM 210J switch
• STM 331 temperature sensor

• People counter