Robust Model-Based Estimators for Cardiac Nerve Activity

Cristian M. Radu Daniel T. Kaplan Department of Physiology and Centre of Nonlinear Dynamics in Physiology and Medicine McGill University, 3655 Drummond, Montr eal H3G-1Y6, Qu ebec, Canada Abstract This paper outlines our design of a continuous estimator for the sympathetic innervation of the heart. The estimator is computed by linear methods, yet it is tested on a nonlinear, detailed model of cardiovascular and respiratory dynamics. Inverting a model output (blood pressure) to recover internal activity by means of the H1 and structured singular value ( ) design methods, allows direct treatment of signal and model uncertainty (noise and nonlinearity, respectively), in a computationally convenient, linear way. Numerical simulations suggest that the variability of the blood pressure signal can be decoded to yield good estimates of nerve activity despite parametric uncertainty, in a range of physiological conditions.