Onboard AI-Driven Autonomous Tasking for Rapid Wildfire Response with a LEO Wide-Swath Observation Satellite

Low Earth Orbit (LEO) remote sensing satellites exhibit significant limitations in emergency response operations, primarily due to the constraints of orbital mechanics and a traditional reliance on ground-involved, open-loop operational models. Current response latencies often span several hours, rendering acquired data less effective for disasters with rapid temporal dynamics. We propose a new intra-pass autonomous reaction application mode to solve this key problem in emergency response. We build the mode on an agile satellite platform operating in an 800 km sun-synchronous orbit. We equip the platform with two core sensors through an innovative design. The two sensors are an optical camera with 0.5-meter resolution and a forward-looking infrared camera. The infrared camera has a 600-kilometer swath and a 10-meter nadir resolution. We take onboard artificial intelligence technologies as the research core. The technologies support real-time and autonomous analysis of forward-looking infrared image data. This analytical capability facilitates the real-time detection and autonomous confirmation of stochastic events, such as wildfires, by leveraging onboard artificial intelligence (AI) to perform real-time analysis of IR imagery. This subsequently triggers immediate autonomous re-tasking for high-resolution follow-up observation within the same orbital pass. We establish an optimization model for the satellite system. The model is used to solve the key balance problem between the sensor forward-looking angle, achievable reaction time, and actual detection resolution. We conduct a detailed engineering time budget analysis for the whole satellite operation system. Analysis results show that a specific forward-looking angle must be set for the infrared camera to obtain a 70-second autonomous response window of the satellite. The 70second window contains 30 seconds for AI cognitive analysis, a maximum of 33.6 seconds for platform attitude maneuvering, and 10 seconds for platform stability adjustment. The minimum forward-looking angle required for the infrared camera is about 33.6 degrees. The ground sampling distance of the infrared camera will decrease to 12.2 meters at this angle. The resulting 12.2-meter ground sampling distance (GSD) remains sufficient to meet the practical requirements of early-stage detection. The proposed Intra-Pass Autonomous Reaction (IPAR) mode reduces emergency response times from hours to seconds, significantly enhancing the proactive observation capabilities of LEO platforms.