An Interpretable Hierarchical Belief Rule Base for the Performance Evaluation of Laser Inertial Measurement Units
Y. N. Gao, Jieyu Liu, Shuanzhu Li, Can Li, Qiang Shen, Zhaoqiang Wang · Entropy · 2026
Laser strapdown inertial measurement units (LIMUs) are critical components of autonomous navigation and attitude determination systems used in mission-critical platforms such as spacecraft. Their performance directly affects the reliability and safety of space missions. To address the challenges posed by high-dimensional indicator sets, limited data, and stringent interpretability requirements, this paper proposes an interpretability-aware hierarchical Bayesian belief rule base model, termed IH-BRB, for LIMU performance evaluation. First, a five-level hierarchical evaluation structure is developed based on the physical architecture and error-propagation relationships of a LIMU, thereby reducing rule base complexity and improving inference traceability. Second, an interlayer transmission mechanism based on the expected grade scores and entropy-derived reliability is introduced to preserve essential uncertainty information and dynamically adjust the attribute weights of upper-level BRB nodes. Third, semantic consistency constraints are imposed on the conditional probability tables and rule-consequent belief distributions to prevent semantic drift during data-driven optimization. The proposed model is validated using an engineering LIMU dataset with expert-defined target grades to assess its consistency with the established grading criterion. The results show that IH-BRB maintains a competitive evaluation performance while preserving physically meaningful rule semantics and layer-wise traceability.