Dynamic Targeting of Satellite Observations Using Supplemental Geostationary Satellite Data and Hierarchical Planning
New hierarchical AI uses 35-minute lookahead data to optimize Earth observation, outperforming traditional methods by up to 41%.
A research team from NASA's Jet Propulsion Laboratory (JPL) and MIT has developed a novel AI planning system that significantly enhances how Earth observation satellites schedule their tasks. The system, detailed in a paper for the 2026 IEEE International Conference on Robotics and Automation, tackles a core challenge in Dynamic Targeting (DT) missions: the limited foresight provided by an onboard 'lookahead' sensor, which typically offers only about one minute of advance data. This short horizon restricts a satellite's ability to intelligently plan observations for complex, dynamic phenomena like cloud formations or emerging storms.
The breakthrough involves a two-tiered, hierarchical planning approach that incorporates a supplemental data stream from geostationary satellites. These high-altitude satellites provide a much broader and longer-term view, offering lookahead information up to 35 minutes in advance. The AI planner first uses this geostationary data to create a high-level, long-term observation 'blueprint' in polynomial time, avoiding an exponential explosion in the search space. It then leverages the satellite's own precise, short-range sensor data to refine and execute the plan. In tests across four scenarios—including three cloud avoidance variations and a storm hunting use case—this method outperformed traditional DT planners relying solely on onboard data by up to 41%, proving most effective when targets are sparsely distributed.
- Integrates geostationary satellite data for a 35-minute planning horizon, a 35x increase over typical 1-minute onboard sensor data.
- Uses a hierarchical AI planner to create long-term blueprints and short-term refinements, solving exponential search space challenges.
- Demonstrated a 41% increase in science return for applications like cloud avoidance and storm tracking compared to previous methods.
Why It Matters
Enables satellites to capture more critical Earth science data, improving weather forecasting, climate monitoring, and disaster response.