Left ventricular assist devices are increasingly used for long-term support in heart failure patients. To promote heart recovery, finding an optimum operating point of the pump that is appropriate for the heart function, and state of the circulatory system is important. Therefore, the baroreflex sensitivity (BRS), which reflect the state of heart function, is used as a control variable. To find the optimum point automatically, an extremum search algorithm (ESA) is designed to find the mean arterial pressure (MAP) that is corresponding to the maximum value of BRS. Then, a MAP controller based on model-free adaptive controller (MFAC) is designed to maintain the measured MAP tracking the desired one. A mathematical model of the cardiovascular system is used to verify the feasibility of the control strategy in the presence of left ventricular (LV) failure, physically active, and a recovery of cardiac function. The simulation shows that the ESA can find the maximum value of BRS automatically. When the peripheral resistance is reduced for simulating a slight physical active, the rotational speed of the pump is automatically increased (6,800 rpm vs. 8,000 rpm). When E(max) is increased from 0.6 to 1.8 mm Hg/ml to mimic a heart recovery, the speed is decreased from 8000 to 7200 rpm, which may avoid the damage of LV contractility. As a key feature, the proposed control strategy finds the optimum operating point of the pump without the need to set reference value of control variable. This feature is benefit for heart recovery.