关键词: 星表巡视器, 非结构化场景, 同步定位与建图, 激光-惯性里程计, 几何退化检测

Abstract: Planetary surface exploration missions impose stringent requirements on the accuracy and reliability of autonomous navigation systems. However, extraterrestrial surfaces (e.g., the Moon and Mars) are characterized by typical unstructured terrains, such as extensive flat sandy areas, irregular craters, and scattered rocks. Such environments often induce geometric degeneracy in tradi-tional LiDAR-Inertial Odometry (LIO) systems along specific degrees of freedom, subsequently causing state estimation diver-gence and trajectory drift, which pose threats to the safe operation of rovers. To address these challenges, this paper proposes a robust SLAM method that integrates online degeneracy detection with an active constraint mechanism. First, during the update phase of the Iterated Error-State Kalman Filter (IESKF), an approximate Fisher Information Matrix (FIM) is constructed using the Hessian matrix derived from the point-to-plane distance observation equation. The observability of the state space is then quantified via real-time eigenvalue decomposition. Second, an active constraint strategy based on eigenvalue thresholds is de-signed. When the eigenvalue in a specific direction falls below a preset threshold, geometric degeneracy is identified. Subse-quently, virtual observations are introduced to impose zero-residual hard constraints, forcing the system to rely on Inertial Meas-urement Unit (IMU) pre-integration constraints in the degenerate dimensions. Experimental results based on the LuSNAR plane-tary simulation dataset demonstrate that, compared with existing LIO algorithms, the proposed method accurately identifies de-generate directions and effectively eliminates unbounded drift in feature-deprived directions, maintaining high trajectory con-sistency during long-distance navigation. The results verify that the proposed active constraint method effectively resolves the geometric degeneracy problem in unstructured planetary environments, significantly enhancing the autonomous localization accuracy and robustness of planetary rovers.

Key words: planetary rover, unstructured environment, simultaneous localization and mapping, LiDAR-inertial odometry, geo-metric degeneracy