Vigi-Forest: Embedded technology protecting endangered forests

Responding to the environmental emergency

Climate change is increasing the risk of wildfires. Rising temperatures and prolonged droughts are making fires more frequent, larger, and more widespread, according to research by INRAE (France’s National Research Institute for Agriculture, Food and Environment). Recent events confirm this trend, notably the wildfires in California at the start of 2025 and, in France, the devastating summer of 2022, when over 30,000 hectares were destroyed in the forests of Gironde.

Connected sensors, first envisioned in the 1990s, have now become technically and economically viable thanks to advances in IoT technologies. These interconnected sensors offer a valuable complement to existing human and visual surveillance methods: they enable rapid detection, cost-effective deployment, and autonomous operation without centralized infrastructure.

Driven by a desire to use our expertise to serve meaningful causes—responsible technologies for the common good, environmental protection, and positive ecological impact—our team of IoT and embedded systems experts has defined the key criteria for a field-ready solution:

• Early detection at the first signs of smoke
• Long-lasting energy autonomy with zero maintenance
• Native interoperability with existing infrastructures
• Environmental respect, with no polluting batteries or toxic waste
• Controlled cost, suitable for large-scale deployment

In this context, open source offers a transparent and modular alternative. With that philosophy in mind, we developed our demonstrator.
 

Introducing Vigi-Forest at Tech & Fest 2025

At the 2025 Tech & Fest exhibition, Smile presented an operational embedded solution capable of detecting early warning signals of wildfires in remote areas in real time.
This project reflects our expertise and ability to integrate software, electronics, and network skills to address a critical environmental challenge: detecting the first signs of a wildfire autonomously, without infrastructure, and under extreme conditions.

Local, autonomous, and proactive detection

Far from traditional infrastructure, forest detection relies on the ability to capture information locally, process it instantly, and transmit it reliably—even in the face of faults or failures. To meet this challenge, we designed an architecture based on the following components:

• Olfactory detection via embedded learning: a BME688 sensor learns to recognize gas signatures typically associated with the onset of a fire using a machine learning model executed directly on the microcontroller.
• Real-time processing: C/C++ code ensures fast, energy-efficient execution, suited to power-constrained environments.
• Transmission via LoRa Mesh: once a threat is detected, the alert is relayed through a self-organizing mesh network. Each node forwards the information toward the farthest point, optimizing hops and bypassing any potential failures.

This architecture enables the transmission of a reliable alert within seconds, without the need for pre-existing infrastructure or continuous human supervision.

LoRa Mesh Network: Autonomy and Continuity

For transmission, we opted for a self-configuring LoRa Mesh network. Each node communicates up to 15 km around it, and the network dynamically adapts to failures to ensure the smooth flow of information, without the need for traditional telecom infrastructure.

Reactive and Interoperable Control Interface

The system is monitored via a custom-built web interface (Symfony + React.js). The goal is to provide a clear, up-to-date, and reliable view of the network's status and alerts, while ensuring full interoperability with third-party sources such as weather or satellite data.

• Dynamic event mapping.
• Multi-sensor aggregation for visual/olfactory/thermal correlation.
• Open APIs to interface with external data in real-time.
   

The Future? Toward an AI-Enhanced Environmental Platform

The integration we've achieved paves the way for architectures capable of processing heterogeneous data streams at the edge—such as video, gas emissions, temperature, or satellite data. Smile has the expertise to design such platforms, combining energy efficiency, embedded intelligence, and network resilience. Integrating the current system with image analysis solutions is also a promising avenue, one that would further enhance overall reliability.

This demonstrator highlights several high-potential technical directions: running AI models on ultra-low-power sensors, designing resilient and self-healing mesh networks, and ensuring seamless interoperability with existing alert systems.

These topics are at the heart of our expertise: distributed architectures, open source software, and embedded system optimization—all essential assets for building IoT solutions that are truly useful, efficient, and suited to real-world constraints.