In recent years, Japan has faced a daunting “dual-threat” profile: the increasing risk of large-scale wildfires and a surge in dangerous bear-human encounters.
According to recent Ministry of the Environment data, the 2025 fiscal year has already seen a record-breaking 13 deaths and nearly 200 injuries from bear attacks nationwide. As bears push deeper into populated areas in search of food, the demand for proactive, automated safeguards has never been more urgent.
Simultaneously, forest fire patterns are becoming more severe. While historical averages typically sat at 1,200 wildfires burning 800 hectares annually, 2025 saw an unprecedented spike, with over 4,400 hectares burned in the first three months alone due to extreme winter dryness.
The Evolution of a Partnership: 2025–2026
In response to these escalating risks, Asteria Corporation (Japan), RoboticsCats (Hong Kong), and Fire Foresight (Australia) have partnered to launch an evolved Proof of Concept (PoC) in Karuizawa that merges AI with IoT technology.
Building on a successful initial pilot project in Hong Kong in early 2025, this initiative represents a 12-month milestone in “Resilience-as-a-Service”.’’
- “Hybrid AI” for Simultaneous Threat Detection
The standout feature of this system is its ability to monitor both wildfire and wildlife risks using a single camera feed:
- AI Wildfire Sentinel: The RoboticsCats engine continuously scans high-resolution feeds, identifying smoke or fire signatures as small as 0.03% of the image to catch outbreaks at the earliest possible stage.
- AI Bear Detection: Utilizing Large Language Model Visual Question Answering (LLM VQA), the system dispatches images to generative AI that can answer complex prompts such as, “Detect bears near structures?” to generate precise alerts.
- Defeating the “Fog” Barrier with Multi-modal Verification
Highland areas like Karuizawa are notoriously prone to fog, which conventional AI often mistakes for smoke—leading to costly false alarms. Our 2026 architecture solves this through Cross-Verification:
- Environmental Intelligence: The Gravio Hubkit edge gateway cross-references visual detections with local air quality sensors measuring PM2.5, TVOC, and CO2.
- Verified Peace of Mind: By using multi-modal data and “human-in-the-loop” validation, the system ensures that alerts are only sent for genuine threats, providing security for residents without unnecessary disruption.
- High Deployability: Maximizing Existing Infrastructure
Another major advantage of this “Smart Resilience” model is that it does not require expensive, dedicated surveillance towers.
- Retrofit Ready: The system can be integrated into existing security networks and CCTV, significantly maximizing ROI for property owners and municipalities.
- Global Track Record: RoboticsCats and Fire Foresight’s combined platforms are already serving customers across 13 countries, ensuring a battle-tested foundation for Japan’s specific needs.
The Path Ahead
As we continue to harden this technology globally, our focus remains on the Wildland-Urban Interface (WUI). We are moving toward adding partner Fire Foresight’s autonomous drone-based response platform for close-range surveillance and rapid fire suppression. We believe that through open innovation and cross-industrial cooperation, we can make “always-on” safety the standard for every community.
Download our solution paper today to learn more about our AI Wildfire Detection Software-as-a-Service.
