Optimal sensor fusion for distributed sensors subject to random delay and packet loss

In this paper we study optimal information fusion for sampled linear systems where the sensors are distributed and measurements are collected to central unit via a wireless network. Every sensor measurement is subject to random delay or might even be completely lost. We show that optimal sensor fusion consist in a time-varying Kalman filter with bufferized measurements. We also propose a suboptimal but computationally efficient fusion architecture based on a bank of static gains that can be optimally designed if packet delay statics are known. Finally, algorithms to check for the existence of stable estimators and to evaluate their error covariance are given and some special cases are analyzed.