A distributed parallel estimation architecture for cooperative vehicle formation control

Parallel estimation architectures have significant advantages in precisely controlled vehicle formations such as space-based high-resolution imaging instruments. Applying standard estimator design procedures to such systems can lead to unanticipated, and potentially destabilizing, disagreement dynamics. These dynamics arise from the noise driven differences between the state estimates in each of the vehicles and are coupled into the full system dynamics by the distributed actuation typical of vehicle formations. We present the parallel estimation problem in terms of the full system dynamics and illustrate both time-varying and time-invariant solutions to the estimator design problem. These solutions are tractable but may not be optimal. A notable feature of the problem is that individual estimators must also propagate the covariances of all of the other estimators in the formation.