Arvin Dehghani

Challenges of adaptive control–past, permanent and future☆

Abstract This paper reviews three different types of challenges to adaptive control. The first group comprises challenges met in the subject’s development. They include difficulties associated with the MIT rule, bursting, the Rohr’s counterexample and unplanned instability in iterative identification and control. An understanding of these phenomena and mitigating strategies are now available. The second group comprises difficulties that are intrinsic to virtually any adaptive control algorithm, and that have frequently been overlooked. For example, if a plant is unknown, and a control objective is set, the objective may in practical terms be unachievable, and any adaptive control algorithm needs to deal with that possibility. The third group comprises some issues to which researchers are currently devoting significant attention, including multiple model adaptive control and model-free design.

Validating Controllers for Internal Stability Utilizing Closed-Loop Data

We introduce novel tests utilizing a limited amount of experimental and possibly noisy data obtained with an existing known stabilizing controller connected to an unknown plant for verifying that the introduction of a proposed new controller will stabilize the plant. The tests depend on the assumption that the unknown plant is stabilized by a known controller and that some knowledge of the closed-loop system, such as noisy frequency response data, is available and on the basis of that knowledge, the use of a new controller appears attractive. The desirability of doing this arises in iterative identification and control algorithms, multiple-model adaptive control, and multi-controller adaptive switching. The proposed tests can be used for SISO and/or MIMO linear time-invariant systems.

Checking if controllers are stabilizing using closed-loop data

Suppose an unknown plant is stabilized by a known controller. Suppose also that some knowledge of the closed-loop system is available and on the basis of that knowledge, the use of a new controller appears attractive, as may arise in iterative control and identification algorithms, and multiple-model adaptive control. The paper presents tests using a limited amount of experimental data obtained with the existing known controller for verifying that introduction of the new controller will stabilize the plant

Verifying stabilizing controllers via closed-loop noisy data: MIMO case

This article introduces novel tests which utilize a limited amount of experimental and possibly noisy data obtained from a stable closed-loop system, i.e. an interconnection of an existing known stabilizing controller and an unknown plant, to infer if the introduction of a prospective controller will stabilize the unknown plant. This extends our earlier results to include the MIMO systems.

HISTORICAL, GENERIC AND CURRENT CHALLENGES OF ADAPTIVE CONTROL

Abstract This paper reviews three different types of challenges to adaptive control. The first group comprises challenges met in the subjects development. They include difficulties associated with the MIT rule, bursting, the Rohrs counterexample and unplanned instability in iterative identification and control. An understanding of these phenomena and mitigating strategies are now available. The second group comprises difficulties that are intrinsic to virtually any adaptive control algorithm, and that have frequently been overlooked. For example, if a plant is unknown, and a control objective is set, the objective may in practical terms be unachievable, and any adaptive control algorithm needs to deal with that possibility. The third group comprises some issues to which researchers are currently devoting significant attention, including multiple model adaptive control and model free design.

Practical novel tests for ensuring safe adaptive control

In this paper, we further illustrate the versatility and effectiveness of our novel tests for ensuring safe adaptive control in practice. The tests utilize a limited amount of experimental and possibly noisy data obtained from a closed-loop-consisting of an existing known stabilizing controller connected to an unknown plant-to infer if the introduction of a prospective controller will stabilize the unknown plant. The need and importance of this arise in iterative identification and control algorithms, multiple-model adaptive control (MMAC), and multi-controller adaptive switching.

A Robust Multiple-Model Adaptive Control System for Automatic Delivery of a Vasoactive Drug

Automatic closed-loop administration of sodium nitroprusside for the regulation of blood pressure in patients experiencing acute hypertension has been the subject of intense research over the last three decades. Yet, to date, manual administration of vasoactive drugs by a human operator remains the standard of care in the clinical setting. This manuscript describes a novel control approach for this application based on Robust Multiple-Model Adaptive Control (RMMAC). The RMMAC architecture features robust controllers designed with μ synthesis and Kalman filtering for system estimation. The new system was coupled with a mathematical model of a patient’s response to drug infusion and tested in computational simulations. The results indicate that the RMMAC approach has the potential to deliver robust performance even in challenging operating conditions, with mean arterial pressure remaining within the specified target range over 99% of the time.

Safety issues in adaptive control systems for automatic administration of vasoactive drugs

Abstract Automatic administration of drugs to control cardiovascular function during and after surgery has received considerable attention. Although the potential benefits associated with such a technology remain unquestioned, the several adaptive control strategies proposed thus far have had very limited success in practice. In developing robust adaptive control methodologies for drug dosing in cardiovascular applications, we have analysed a well-known multiple-model adaptive control strategy for blood pressure control. The results reveal that no guarantee of protection against actually inserting a destabilising controller into the closed-loop is given and one cannot even put a global upper bound on the time during which the destabilising controller is attached. We advocate caution towards issues which in the past may have been either disregarded or not subjected to a systematic analysis as instability could be fatal in the context of a clinical application.

Verifying closed-loop performance before inserting a new controller

Consider a closed-loop interconnection of an unknown linear plant and a known linear stabilizing controller, and assume that some knowledge of the closed-loop system is available. Suppose-on the basis of that knowledge-the existing, still stabilizing, controller no longer provides a satisfactory closed-loop performance, and hence the use of a new controller appears attractive. This scenario is common in Multiple Model Adaptive Control and Iterative Identification and Control algorithms. Our results will provide assurance for the closed-loop performance before the insertion of a new controller, and complement our earlier results for guaranteeing closed-loop stability in advance.

Verifying nonlinear controllers for stability utilizing closed-loop noisy data

A framework for addressing a potential instability problem in adaptive control and iterative identification and controller design algorithms is proposed. Suppose an unknown plant is stabilized by a known controller, some knowledge of this stable closed-loop system is available, and the use of a new controller to replace the current stabilizing controller becomes imminent. Our analysis results assume that the ‘unknown’ plant and the controllers are all nonlinear. We further develop a databased test which utilizes a limited amount of experimental data from an existing stable closed-loop (a plant in connection with a linear stabilizing controller) for verifying that the introduction of a new nonlinear controller will stabilize the unknown plant.