基于数值微分技术构造鲁棒估计模型的方法及其应用仿真

Constructing the Robust Estimation Models by Numerical Differentiation and the Simulations for Their Applications

在研究宏观空间运动特性的基础上,利用数值微分作为表达工具描述未知运动过程的动态特性,构造了数值微分型滤波模型(NDFM)和数值微分型滤波-预报联合模型(NDFPM)。这种方法能根据各种应用要求对动态特性完全未知的运动过程建立结构简单、鲁棒性强的估计模型,而且容易选择估计算法获得满意的性能。本文对未知扰动作用下的被控过程建立NDFM并实现状态重构和扰动补偿;对动态未知的被跟踪目标建立NDFPM并估计出运动参数的当前值和一步预报值。仿真结果表明这两种模型具有较强的鲁棒性和满意的估计精度。

Researching the characters of macro motion, numerical differentiation is introduced to describe dynamics of unknown kinematics process, then the filtering model of numerical differentiation (NDFM) and the combined filtering-predicting model of numerical differentiation (NDFPM) are constructed. Even though the dynamics of the process to be estimated is unknown, robust and simply models can be created by this approach for various applications, and it is easy to select appropriate estimation algorithms for satisfied estimation quality. The states of control-system disturbed by unknown inputs are restructured by NDFM for feedback and compensating disturbances; the present-predictive estimation of kinematics parameters are captured by NDFPM for tracked object with unknown dynamics. Simulation results show that NDFM and NDFPM are robust models to obtain precise estimation.