Heat Intelligence at the Edge

• Measures air temperature, relative humidity, radiant heat and air movement, and computes real heat stress indices such as heat index and WBGT locally, rather than reporting temperature alone
• Predicts how hot a specific indoor space will get over the next few hours, learning the thermal behaviour of that particular room from its own history
• Grades risk against the person and the activity, so a labourer at work, a child in a classroom and an elderly person at home receive different thresholds and different advice
• Triggers local relief automatically — switching on a fan or cooler, opening ventilation, starting evaporative cooling, or sounding a rest-break alert on a worksite
• Continues to function during power cuts and network outages, since these are most likely exactly when heat risk peaks
• Builds a street-level heat map from multiple nodes that a municipality can use to decide where to plant trees, apply cool roofs, install shade or place water points

Mandatory hardware requirement: The solution must be built on the Arduino UNO Q. All AI inference must run on the board itself. Cloud APIs and external AI services may not be used for core decision-making. Sensors, actuators, displays and communication modules are your choice, but the UNO Q must be the computing platform of the submitted prototype.

Technical feasibility (the cheat sheet): Board — UNO Q's Qualcomm Dragonwing QRB2210 Linux processor runs the prediction models while the STM32U585 MCU handles low-power sensing and relay control. Sensors — SHT41 or BME688 for temperature and humidity, DS18B20 inside a matte black globe for radiant temperature, an anemometer or hot-wire sensor for air movement, MLX90640 thermal array for wall and roof surface temperature, and a UV or lux sensor for solar load. Models — regression for WBGT estimation, a compact LSTM or TCN for indoor temperature forecasting, anomaly detection for early heat stress warning, all quantised and run through TensorFlow Lite on the Cortex-A53 cores. Software — Arduino App Lab, Python for the model, C++ on the MCU for actuation, alerts and sleep scheduling. Actuation and alerts — relay-driven fan or cooler control, local buzzer and display, plus SMS or app notification when a link is available. Power — solar panel with a LiFePO4 battery so the node survives long outages, deep sleep managed on the MCU side.

For the Grand Finale, your build should include:

• A working UNO Q node computing local heat stress indices (heat index, WBGT) from real sensor input, not just raw temperature
• An on-device room-level thermal forecasting model, with evidence it learns that specific room's behaviour
• Person-and-activity risk grading logic with at least three distinct profiles and their thresholds documented
• A live automatic relief loop — sensing to decision to actuation — demonstrated end to end with the network and mains disconnected
• A multi-node street-level heat map output plus a one-page civic recommendation showing how a municipality would use it

The Grand Finale is judged on:

• On-device heat intelligence and offline/outage resilience — 30%
• Accuracy of heat stress computation and room-level forecasting — 25%
• Personalised risk grading and the automatic relief loop — 20%
• Multi-node heat mapping and usefulness to a Heat Action Plan — 15%
• Demo quality and clarity of storytelling — 10%