A centralized multi-objective model predictive control for a biventricular assist device: An in vitro evaluation

Abstract

Control of a biventricular assist device (BiVAD) is more challenging than control of a left ventricular assist device due to the process interactions between control loops in a multi-input-multi-output system. Hence, a single centralized multi-objective model predictive controller (CMO-MPC) has been developed to control a BiVAD. The CMO-MPC aims to: 1) adapt pump flow rate according to the Frank-Starling mechanism, 2) avoid ventricular suction, and 3) avoid vascular congestion. The CMO-MPC was benchmarked against a constant-speed (CS) setting in exercise, postural change, and systemic vascular resistance change tests in a mock circulation loop. The CMO-MPC increased pump flow rate from 5.0 L/min to 7.6 L/min in the exercise scenario, which was higher than the pump flow rate in the CS setting (6.0 L/min). In the postural change test, right ventricular end diastolic pressure (RVEDP) decreased to a minimum at 0.1 mmHg and 2.0 mmHg in the CS setting and the CMO-MPC, respectively, indicating that the CMO-MPC could minimize the risk of ventricular suction (with higher minimum RVEDP than the CS setting) when there was a sudden decrease in venous return. In all tests, the CMO-MPC could adapt pump flow rate without resulting events of ventricular suction and vascular congestion.