摘要
针对多源导航信息融合性能评估过程中指标体系维度低、评估方法主观依赖性强以及权重分配机制失准等关键问题,提出一种基于投影寻踪模型的动态评估方法。首先,基于多源导航信息融合过程机理,构建涵盖准确性、实时性、鲁棒性及可靠性的多层级指标体系。其次,设计基于类内密度-类间距离联合优化的投影寻踪模型,通过非线性投影函数重构指标间高阶交互关系,克服传统线性加权方法的信息截断与主观偏倚问题。最后,提出模拟退火驱动的自适应投影方向优化机制,从而规避高维非凸目标函数的局部收敛风险,确保投影方向全局最优性。仿真结果表明,相比于准则重要性评价法、层次分析法和逼近理想解排序法,所提方法的斯皮尔曼系数分别提升15.97%、11.29%与11.94%,有效提高了评估体系的普适性。
Aiming to address key issues likethe low dimension ofthe index system,the strong subjective dependence of the evaluation method,and inaccurate weight distribution in multi-source navigation information fusion performance evaluation,a dynamic evaluation method based on theprojection pursuit modelis proposed.First,based on the processmechanism of multi-source navigation fusion,a multi-level index system covering accuracy,real-time performance,robustness,and reliabilityis constructed.Next,a projection pursuit model with joint optimization of intra-class density and inter-class distanceis designed,using a nonlinear projection function to reconstruct the high-order interaction among indicators,overcoming the information truncation and subjective bias issues inherent in traditional linear weighting methods.Finally,an adaptive projectiondirection optimization mechanism driven by simulated annealing is proposedto avoid the local convergence risk of high-dimensional non-convex functions and ensure global optimality.Simulation results show that,compared with the criteria importance through intercrieria correlationmethod,analytic hierarchy process,and technique fororder preference by similarity to an ideal solution,the Spearman coefficient of the proposed method increases by 15.97%,11.29%and11.94%,respectively,significantly improving the universality of the evaluation system.
作者
白靖
徐博
刘猛
王永召
王朝阳
刘伟
BAI Jing;XU Bo;LIU Meng;WANG Yongzhao;WANG Zhaoyang;LIU Wei(Schoolof Intelligent Science and Engineering,Harbin Engineering University,Harbin 150001,China;Tianjin Navigation Instrument Research Institute,Tianjin 300131,China;CSSC Marine TechnologyCO.,LTD.,Shanghai200136,China)
出处
《中国惯性技术学报》
北大核心
2025年第12期1253-1262,1272,共11页
Journal of Chinese Inertial Technology
基金
国家自然科学基金项目(62473106)
海南省重点研发项目(ZDYF2024GXJS009)
海洋防务基金(JJ-2023-705-01)
航空科学基金(2024Z0710P6001)。
关键词
多源导航
性能评估
信息融合
投影寻踪
multi-source navigation
performance evaluation
information fusion
projection pursuit
