视觉-LiDAR 融合的空间目标悬停段相对运动 参数自适应滤波估计

To meet the demand for high-precision and continuous state estimation of non-cooperative targets in on-orbit servicing missions，and to overcome the limitations of single optical or laser sensors in depth perception and environmental adaptability，an adaptive filtering method for motion parameter estimation during the hovering phase is proposed，which fuses stereo vision and LiDAR （light detection and ranging） information. On the basis of establishing the relative kinematic model，the inertia ratio logarithmic parameterization is adopted to handle the uncertainty of the target inertia matrix，and an adaptive error-state Kalman filter（ESKF）algorithm is designed. An adaptive mechanism is introduced to cope with time-varying noise，the observation noise covariance is recursively adjusted in real time using innovation statistics，and a regularized projection method based on eigenvalue decomposition is combined to ensure the positive definiteness of the covariance matrix，thereby enhancing the adaptability and stability of the filter. The Cramér-Rao lower bound （CRLB） is derived through theoretical analysis，which proves that the estimation error variance of the fusion scheme is superior to that of the single-sensor scheme. Numerical simulation results show that compared with the single-sensor measurement method，the proposed algorithm significantly improves the estimation accuracy of key indicators such as attitude，angular velocity and feature point position.