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

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.