IoT and Edge AI

What are some examples of IoT projects using embedded systems?

Twelve IoT project examples built on embedded systems, from a Wi-Fi weather station to a LoRa sensor and a TinyML device, with hardware and skills for each.

On this page
  1. Quick answer: 12 example projects
  2. Comparison at a glance
  3. Beginner projects
  4. Intermediate projects
  5. Advanced projects
  6. What every IoT project shares
  7. How to choose your first project
  8. Safety notes
  9. FAQ

Quick answer: 12 example projects

IoT projects using embedded systems pair a microcontroller and sensors with a network connection, so the device can report data or be controlled remotely. Good examples, roughly from easiest to hardest:

  1. Wi-Fi weather station that logs temperature, humidity and pressure to a dashboard
  2. Plant soil-moisture monitor that alerts you when it’s time to water
  3. Smart relay or plug controller switched from a phone or home automation system
  4. Bluetooth LE environmental beacon readable from a phone
  5. Battery-powered door or window sensor that lasts months on a small battery
  6. Home energy monitor using clamp-on current sensors
  7. LoRa long-range sensor node for gardens, farms or remote sheds
  8. Air-quality monitor measuring particulates and CO₂
  9. GPS asset tracker reporting location over cellular IoT
  10. Voice keyword or gesture recognition device using TinyML
  11. Vibration monitor for predictive maintenance on a motor or pump
  12. Zigbee, Thread or Matter device that joins a commercial smart home system

Every one of these has a real-world counterpart in commercial products, which is what makes them useful practice. Each one teaches a different part of embedded IoT engineering.

Comparison at a glance

Project Typical MCU Connectivity Difficulty Main skill it teaches
Weather station ESP32 Wi-Fi + MQTT Beginner I2C sensors, publishing data
Soil-moisture monitor ESP32 or ESP8266 Wi-Fi Beginner ADC readings, calibration, deep sleep
Smart relay ESP32 Wi-Fi + MQTT Beginner Remote control, subscribing to commands
BLE beacon nRF52/nRF54, ESP32 Bluetooth LE Beginner–intermediate BLE advertising, GATT
Door/window sensor nRF52, ESP32-C6, low-power MCU BLE, Zigbee or Thread Intermediate Interrupt wake-up, power budgeting
Energy monitor ESP32 Wi-Fi Intermediate Analog signal sampling, RMS calculation
LoRa sensor node STM32WL, ESP32 + SX126x radio LoRa / LoRaWAN Intermediate Long-range radio, tiny payloads, duty cycle
Air-quality monitor ESP32 Wi-Fi Intermediate UART/I2C sensors, data smoothing
GPS asset tracker nRF91 series or MCU + cellular modem LTE-M / NB-IoT Advanced Cellular IoT, GNSS, power management
TinyML keyword/gesture Nordic nRF52840, ESP32-S3, Cortex-M4F/M7 BLE or Wi-Fi Advanced On-device ML, DSP, memory budgeting
Vibration monitor ESP32-S3, STM32 Wi-Fi or LoRa Advanced High-rate sampling, FFT, anomaly detection
Zigbee/Thread/Matter device ESP32-C6/H2, nRF52840/nRF54, Silicon Labs EFR32 802.15.4 mesh Advanced Standard device models, commissioning

Beginner projects

Wi-Fi weather station

What it does: Reads temperature, humidity and air pressure every few minutes and sends them to a dashboard.

Typical hardware: An ESP32 development board and an environmental sensor such as a BME280 on I2C.

How it works: The firmware connects to Wi-Fi, reads the sensor, and publishes the values to an MQTT broker. A tool like Home Assistant, Node-RED or Grafana graphs the data. Our ESP32 MQTT tutorial covers exactly this pattern.

What you learn: I2C, sensor libraries, Wi-Fi connection handling, MQTT topics and payload formats such as JSON.

Real-world equivalent: Commercial weather stations and building environmental sensors.

Plant soil-moisture monitor

What it does: Measures soil moisture and sends an alert, or switches on a small pump, when the soil gets too dry.

Typical hardware: ESP32 or ESP8266, a capacitive soil-moisture sensor (these usually last longer than the resistive type, which corrode), and optionally a relay and pump.

What you learn: Reading the ADC, calibrating a sensor against “dry” and “wet” reference values, and using deep sleep to run on batteries by waking only every so often.

Real-world equivalent: Precision agriculture sensors and smart irrigation controllers.

Smart relay or plug controller

What it does: Switches a lamp or fan on and off from a phone app or a home automation system, and reports its state.

Typical hardware: ESP32 with a relay module. For anything mains-powered, a ready-made smart plug that can be reflashed with open firmware such as ESPHome or Tasmota is a much safer route than wiring your own (see the safety notes below).

What you learn: Subscribing to command topics, keeping device state consistent after reboots, and handling reconnections.

Real-world equivalent: Smart plugs, switches and lighting controllers.

Bluetooth LE environmental beacon

What it does: Broadcasts temperature and battery level over Bluetooth LE so a phone or gateway can read them without pairing.

Typical hardware: A Nordic nRF52-series board or an ESP32.

What you learn: BLE advertising packets, the GATT services and characteristics model, and how small radio choices affect battery life.

Real-world equivalent: Asset tags, cold-chain loggers and fitness sensors.

Intermediate projects

Battery-powered door or window sensor

What it does: Reports when a door opens or closes, running for months or longer on a coin cell or small battery.

Typical hardware: A reed switch and magnet, with a low-power MCU such as an nRF52, an ESP32-C6, or a Zigbee/Thread chip.

What you learn: Waking from deep sleep on a pin interrupt, debouncing, and power budgeting: measuring sleep current and transmission bursts to estimate battery life. Wi-Fi is usually a poor fit here because of its connection overhead, which is a useful lesson in choosing a wireless link.

Real-world equivalent: Alarm system contacts and smart home sensors.

Home energy monitor

What it does: Measures electricity use on circuits and shows live and historical consumption.

Typical hardware: ESP32 with non-invasive split-core current transformer (CT) clamps that clip around a cable, plus burden resistors and biasing for the ADC. Dedicated energy-metering ICs are another option.

What you learn: Sampling AC signals, calculating RMS current and real power, and filtering noisy analog data.

Real-world equivalent: Smart meters and building energy management systems.

Note: Installing clamps inside a consumer unit or breaker panel means working near live conductors. In many places that should be done by a qualified electrician.

LoRa long-range sensor node

What it does: Sends small sensor readings over distances that Wi-Fi can’t reach, such as across a farm or from a remote outbuilding.

Typical hardware: An STM32WL (MCU and LoRa radio in one chip), or an ESP32 or RP2040 paired with a Semtech SX126x-based radio module. Use LoRaWAN through a community or private network server, or plain LoRa point-to-point.

What you learn: Packing data into a few bytes, regional frequency rules and duty-cycle limits, and the difference between LoRa (the radio modulation) and LoRaWAN (the network protocol).

Real-world equivalent: Agricultural sensors, water-level monitors and smart city parking sensors.

Air-quality monitor

What it does: Measures particulate matter (PM2.5/PM10), CO₂ and possibly VOCs, and displays or uploads the readings.

Typical hardware: ESP32 with a laser particulate sensor (usually over UART) and an NDIR CO₂ sensor such as the Sensirion SCD4x series (I2C).

What you learn: Handling several sensors with different interfaces, warm-up times and sampling rates, plus smoothing readings and adding a small display.

Real-world equivalent: Indoor air-quality monitors and HVAC demand-controlled ventilation.

Advanced projects

GPS asset tracker over cellular

What it does: Reports a vehicle’s, bike’s or shipment’s location periodically over a cellular network.

Typical hardware: A Nordic nRF91-series board, which combines an MCU, an LTE-M/NB-IoT modem and GNSS, or an MCU paired with a cellular modem module. You’ll need a SIM with an IoT data plan.

What you learn: Cellular power-saving modes (PSM and eDRX), GNSS fix times, handling poor coverage, and data costs.

Real-world equivalent: Fleet telematics, container tracking and bike-share locks.

Voice keyword or gesture recognition with TinyML

What it does: Recognises a spoken keyword or a hand gesture on the device itself, and sends only the result over the network.

Typical hardware: Boards with microphones and IMUs, such as Arduino Nano 33 BLE Sense-class boards (nRF52840), ESP32-S3 boards, or STM32 discovery kits.

What you learn: Collecting training data, feature extraction (spectrograms for audio, windows of accelerometer data for motion), quantized models, and fitting inference into limited RAM. Start with Getting Started With TensorFlow Lite for Microcontrollers, and see Embedded AI and Computer Vision at the Edge for the bigger picture.

Real-world equivalent: Wake-word detection in smart speakers and gesture control in wearables.

Vibration monitor for predictive maintenance

What it does: Watches a motor, pump or fan for changes in its vibration pattern that could signal wear.

Typical hardware: An MCU with a fast accelerometer on SPI, such as an ESP32-S3 or STM32, sending summaries over Wi-Fi or LoRa.

What you learn: High-rate sampling with DMA, FFTs, extracting features like RMS and peak frequencies, and sending summaries instead of raw data. You can add an anomaly-detection model as a stretch goal.

Real-world equivalent: Industrial condition-monitoring sensors.

Zigbee, Thread or Matter device

What it does: Builds a light, sensor or switch that joins a commercial smart home hub alongside off-the-shelf products.

Typical hardware: IEEE 802.15.4-capable chips such as the ESP32-C6/H2, Nordic nRF52840/nRF54 series or Silicon Labs EFR32, using the vendor’s Zigbee, Thread or Matter SDK.

What you learn: Standard device data models (clusters), commissioning and pairing flows, mesh networking and interoperability testing. Note that selling a product with Matter or Zigbee branding requires certification. Hobby builds don’t.

Real-world equivalent: Nearly every modern smart home sensor and light.

What every IoT project shares

Under the surface, most connected-device firmware follows the same skeleton:

  1. Initialise hardware, sensors and storage
  2. Connect to the network, with retries and back-off if it fails
  3. Sample sensors on a schedule or on an interrupt
  4. Process locally: filter, check thresholds, maybe run a model
  5. Publish data in a compact, well-defined format
  6. Listen for commands and configuration changes
  7. Sleep to save power, or wait for the next event
  8. Recover from errors with a watchdog timer, and reconnect after outages
  9. Update firmware remotely, ideally, using over-the-air updates

Getting steps 2, 8 and 9 right is what separates a demo that works on your desk from a device that keeps working for a year. To see how these devices fit into the bigger system of gateways, cloud and apps, read IoT System Architecture.

How to choose your first project

  • Pick something you’ll actually use. A sensor in your own home or garden gets maintained and improved.
  • Start with Wi-Fi and an ESP32 if you’re new. The tooling, examples and community support are excellent.
  • Add one new skill at a time: battery power, then a new radio, then on-device ML.
  • Plan the data’s destination first. Decide where readings will go (a dashboard, Home Assistant, a spreadsheet) before writing firmware.
  • Measure power early if the device will run on batteries.

Safety notes

  • Mains electricity can kill. Projects that switch or measure mains voltage need proper isolation, enclosures and fusing, and in many places they need to be installed by a qualified person. Pre-certified smart plugs or low-voltage loads are safer ways to learn.
  • Lithium batteries need appropriate protection circuits and chargers. Damaged or poorly charged cells can catch fire.
  • Radio rules vary by country. Use the correct frequency band and power limits for your region, particularly for LoRa.

FAQ

Which microcontroller is best for IoT projects? For beginners, the ESP32 family is the most common choice because it includes Wi-Fi and Bluetooth, is inexpensive, and works with both the Arduino framework and Espressif’s ESP-IDF. Nordic’s nRF series is popular for low-power Bluetooth devices, and the STM32WL for LoRa.

Can I build IoT projects with a Raspberry Pi? Yes, but a Raspberry Pi is a Linux computer rather than a microcontroller. It’s great as a gateway, broker or dashboard server. For battery-powered sensor nodes, a microcontroller is usually a better fit. The Raspberry Pi Pico W is a microcontroller board with Wi-Fi.

What’s the difference between an embedded project and an IoT project? An embedded project can work entirely on its own, like a thermostat with a display. It becomes an IoT project once it connects to a network to share data or accept remote control. See Embedded Systems and the Internet of Things for the full comparison.

Are these good final-year engineering projects? Several are, especially the advanced ones. Projects that combine sensing, connectivity, power optimisation and a measurable result (like battery life or detection accuracy) tend to make stronger reports than ones that only display data.

The complete guide to iot and edge ai

Embedded Systems and the Internet of Things
Read the guide

Keep exploring

Guides from other topics

All topics →