To address the nonlinear orbit determination problem under multi-satellite cooperative observation, this paper proposes an orbit determination method integrating a plane-constrained observation model with adaptive robust filtering. Based on angular measurements from multiple observation nodes, a linearized observation model is constructed using spatial geometric constraints. The Maximum Correntropy Criterion is then introduced to adaptively weight each measurement component, and a hybrid kernel function is employed to suppress the effects of non-Gaussian noise and outliers. Meanwhile, an adaptive factor based on the covariance matching principle is designed to adjust the process noise intensity online, thereby improving the robustness of the Cubature Kalman Filter in state prediction and update. Simulation results under severe non-Gaussian noise show that the proposed adaptive robust cubature Kalman filter (ARCKF) reduces the position RMSE from 95.3 m for CKF to 30.8 m, corresponding to an improvement of approximately 67.7%, while increasing the computation time from 6.52 s to 7.35 s. These results indicate that the proposed method can achieve improved accuracy and robustness under uncertain measurement statistics and dynamic disturbances, making it suitable for space-based angles-only orbit determination, although further computational optimization is still required for onboard applications.
Li et al. (2026) studied this question.