Adaptive control of arterial blood pressure with a learning controller based on multilayer neural networks

We discuss a two-model multilayer neural network controller for adaptive control of mean arterial blood pressure (MABP) using sodium nitroprusside. A model with an autoregressive moving average (ARMA), representing the dynamics of the system, and a modified back-propagation training algorithm are used to design the control system to meet specified objectives of design (settling time and undershoot/overshoot) and clinical constraints. The controller is associated with a weighting-determinant unit (WDU) to determine and update the output weighting factor of the parallel two-model neural network for adequate control action and a control-signal modification unit (CMU) to comply with clinical constraints and to suppress the effect of adverse noise and to improve the WDU performance. Extensive computer simulations indicate satisfactory performance and robustness of the proposed controller in the presence of much noise, over the full range of plant parameters, uncertainties, and large variations of parameters.