“…One adverse consequence of these unstable estimation and control schemes is that they end up taking longer to converge compared to stable schemes with the same initial conditions and same initial transient behavior. Attitude observers and filtering schemes on SO(3) and SE(3) have been reported in, e.g., [4], [14], [15], [17], [18], [19], [24], [27], [31], [32]. These estimators do not suffer from kinematic singularities like estimators using coordinate descriptions of attitude, and they do not suffer from the unstable unwinding phenomenon encountered by continuous estimators using unit quaternions.…”