Data-Driven Approach for Satellite Onboard Observation Task Planning Based on Ensemble Learning

The chapter introduces a data-driven approach for satellite onboard observation task planning based on ensemble learning. Traditional methods rely on heuristic search algorithms, but this approach leverages machine learning to optimize task scheduling without the need for such algorithms. The proposed framework consists of offline learning and online decision-making phases, where a classifier is trained to decide whether an observation task should be scheduled. Five types of features are designed to represent the observation tasks, and three ensemble learning methods—Extreme Gradient Boosting (XGBoost), Gradient Boosting Decision Tree (GBDT), and Random Forest (RF)—are used to train the decision-maker. Experimental results demonstrate that this approach achieves smaller profit gaps and shorter response times compared to traditional heuristic methods, highlighting the potential of machine learning in optimizing complex task planning problems in the aerospace industry.