Adaptive Control of Arterial Pressure: A Supervisor can Improve Safety and Efficacy

Abstract

The use of adaptive control to regulate the infusion of sodium nitroprusside (SNP) for the control of mean arterial pressure (MAP) is not a new concept. Many investigators have applied various control schemes to this problem. These include: adaptive proportional-integral-derivative (PID) control [1,2], rule based control [3], self-tuning regulators (STR) [4,5], model-reference adaptive systems (MRAS) [6,7], multiplemodel adaptive control (MMAC) [8], and fuzzy control [9]. Most investigators have demonstrated success through simulations and animal experiments, but few have reported clinical use [1,3]. A major consideration in this situation is that success has usually been defined as meeting specified step-response criteria [2]. A controller designed to meet these criteria, considering the possible infusion delays and SNP half-life, has to be fairly aggressive. An aggressive controller can react improperly to the large cardiovascular disturbances that are common in the clinical environment. A conservative controller designed, using standard control design techniques, to be insensitive to the large disturbances tends to be sluggish and unable to attain the desired settling times. An alternative approach is to design an aggressive controller, using standard techniques, to achieve the desired settling times and then design around the controller a SUPERVISOR that can act as a safety net, modifying the controller function when large disturbances are detected.