An adaptive blood pressure control system based on the long-range or finite-time-horizon prediction strategy known as generalized predictive control has been examined. The system was designed to run as a real-time, multitasking operating system providing real-time monitoring and control of mean arterial pressure. The ability to control in the presence of constraints was also incorporated into the system.
Results from both unconstrained and constrained control runs were obtained from experiments on aneasthetized dogs. The problems caused by model-plant mismatch, varying time delays and nonlinearity were handled satisfactorily by the system. The disturbance rejection of the constrained algorithm is better than the unconstrained one. The results demonstrate the applicability of both the constrained and unconstrained generalized predictive control algorithm for regulation of mean arterial pressure. KEY WORDS Mean arterial pressure Predictive control Self-tuning control 1. INTRODUCTION Methods for the automated control of mean arterial pressure (MAP) in post-operative patients have been investigated by many researchers. Some control systems based on feedback control theory including constraints have also been put into trials or routine clinical practice. -' However, improvements in the robustness of the closed-loop control algorithm and the system's adaptivity to a wide variety of patients of differing sensitivities are required. These changes are necessary because patients' response to commonly used vasodilators, e.g. sodium nitroprusside (SNP), are different and the response may even change within the same patient during the course of continuous drug infusion. Therefore systems having the ability of adapting to changes on-line are more desirable for arterial pressure control. Many closed-loop adaptive control systems have thus been developed and tested. A number of them are of the self-tuning type. A (non-adaptive) minimum variance controller has been implemented by Koivo and c o -~o r k e r s . ~. ~ Meline et al.6 describe the use of a self-tuning controller with a minimum variance control law. In order to combat the varying and unknown time delay problem, a pole-placement-based self-tuning method was attempted in a multipledrug infusion system by Serna et al. ' Similar systems in the form of the generalized minimum variance controller were tested by Arnsparger et al., Stern et al.