Abstract
Multi-variable adaptive control of a left ventricular assist pump was studied. For the estimation of the system parameters, second order SISO AR (single input- single output autoregressive) models were introduced. The controlled variable is mean arterial pressure (mAoP), mean atrial pressure (mLAP), or mean blood flow (mBF) and the control input is vacuum pressure (VP) of the pump. Recursive least-squares method with exponential forgetting and covariance reset is used to estimate the parameters. Based on the estimated parameters, controlled variables in steady state for given vacuum pressures are predicted, and the desired vacuum pressure minimizing a performance index is searched. The performance index is weighed summation of square errors in the steady state.
Feasibility of the parameter estimator was studied by computer simulation for the various changes in the circulatory condition: decrease of left ventricular contractility at various rates, change in peripheral resistance, or arhythmia. The controller was able to keep the controlled variable in desired level. Then one input- two outputs (mAoP and mLAP) contoller was studied. In one example, left ventricular contractility suddenly decreased as much as the natural heart ejected no flow, and additionally peripheral resistance increased by 5% at the same time. mAoP dropped from 115mmHg to 97.5mmHg and mLAP rose from 3mmHg to 4.2mmHg instantaneously. But the controller smoothly restored the desired state (mAoP of 114mmHg, mLAP of 3mmHg) after 30 beats. The developed controller adaptively controlled two variables; mean arterial pressure and atrial pressure, for the various changes in the circulatory system. The estimator and controller is feasible for adaptive control of the left ventricular assist heart.
