ABSTRACT Precise on‐orbit identification of solar radiation pressure is essential for space‐based gravitational wave detection spacecraft to enter science mode and achieve mission objectives. However, separating subtle environmental variations from non‐stationary system noise remains challenging. This letter proposes a robust identification framework that integrates deep multi‐period feature extraction with density‐based outlier rejection. Validated using a dataset generated by a high‐fidelity gravitational wave spacecraft formation flight simulation system, the proposed method achieves a vector identification error of 1.63% under an optimal observation window of , demonstrating superior performance and robustness for drag‐free control applications.
Guo et al. (2026) studied this question.