两种最优观测融合方法的功能等价性

On functional equivalence of two measurement fusion methods

中文关键词  多传感器数据融合  集中式观测融合  加权观测融合  Kalman滤波  功能等价性

英文关键词  multisensor data fusion  centralized measurement fusion  weighted measurement fusion  Kalman filtering  functional equivalence

中文摘要

      对于基于Kalman滤波的多传感器数据融合,有两种最优观测融合方法:第一种是集中式观测融合方法,它是通过增加观测向量的维数合并多传感器数据,而第二种是分布式观测融合方法,它是在线性最小方差准则下,通过加权合并多传感器数据,但观测向量维数不变.在数据融合所用的传感器带有相同观测阵的情形下,本文用Kalman证明了两种观测融合方法是完全功能等价的,即用两种方法得到的Kalman估值器(滤波器,预报器,平滑器),信号估值器和白噪声估值器分别在数值上是相等的.在这种情形下,第二种方法不仅可给出像第一种方法一样的全局最优融合估计,而且可明显减小计算负担,便于实时应用.一个数值例子说明了其正确性.

英文摘要

      Currently there exist two optimal measurement fusion methods for Kalman filtering-based multi-sensor data fusion.The first is the centralized measurement fusion method,which combines the multi-sensor data by increasing the dimension of the measurement vector,whereas the second is the distributed measurement fusion method which combines the multi-sensor data by the weighting based on a linear minimum variance criterion,but the dimension of the measurement vector is not changed.By the Kalman filtering method,this paper shows that the two measurement fusion methods are completely functionally equivalent if the sensors used for data fusion have identical measurement matrices,i.e.the Kalman estimators(filter,predictor,smoother),signal estimators,and white noise estimators obtained by two methods are numerically equal,respectively.In this case,the second method not only gives the globally optimal fused estimation as given by the first method,but also obviously reduces the computational burden for real time applications.Finally,a numerical example shows its validity.