一类多速率动态系统的异步数据融合算法

Asynchronous data fusion algorithm based on a class of multirate dynamic systems

针对不同传感器以不同采样率、异步对同一目标进行观测的一类线性时不变动态系统,给出了一种有效的状态融合估计方法。利用该方法进行状态估计,首先根据多尺度系统理论,针对每一个传感器分别建立起相应的系统模型;然后利用Kalman滤波和有反馈分布式融合结构进行数据融合并给出状态估计。该方法避免了插值以及状态和观测的扩维,具有较好的实时性。理论分析和仿真结果均表明,融合估计结果在估计误差方差最小意义下,优于最高采样率的传感器Kalman滤波的结果,融合算法是有效的。

An asynchronous data fusion algorithm for a class of linear time-invariant dynamic systems was presented. There were multiple sensors observing the same single target with different sampling rates asynchronously. Firstly,based on multiscale system theory,the system models were established at each coarse scale aimed at each sensor that had lower sampling rates. The states that the sensors with lower sampling rates observed at coarse scales were modeled as the states average at the finest scale of a proper period approximately with the system noises being omitted. The observations of different sensors at different scales were connected with the state at the highest sampling rate. Secondly,the fused state estimation was obtained using Kalman filter and the distributed structure with feedback. The proposed method could avoide the interpolation and augmentation of state or measurement dimensions,and had a good real time property. The measurements with lower sampling rates were used to estimate the state at coarse scales,while the state estimations were regressed to the finest scale and used to update the state estimation at the finest scale. Theoretical analysis and simulation results show that the fused estimation is better than the Kalman filter result of the sensor with the highest sampling rate,and the algorithm is effective.