A Scheduling Approach for Systems with Actuator Constraints

University Of California-Irvine, Irvine CA

Investigators

Abstract

A Scheduling Approach for Systems with Actuator Constraints CMS-0510874 PI: F. Jabbari University of California-Irvine Abstract This research seeks to develop a new class of high performance output-feedback scheduled controllers, for systems with bounded actuators and other constraints (such as actuator rate or state bounds). Performance is defined as guaranteed levels of disturbance attenuation, primarily in terms of the energy gain from the disturbance to the outputs, though other measures are also possible. The main assumptions are knowledge of the constraints (e.g., saturation limits, rate bounds) and a possibly conservative estimate of the worst-case disturbance (e.g., maximum peak ground acceleration in earthquake applications). The estimate for the peak disturbance is used to establish stability and minimum performance guarantees in all cases. The scheduling is aimed at removing potential conservatism, and providing stronger performance, when the disturbance is less severe than the worst case. The proposed research includes control theoretic tasks, aimed at characterizing the scheduled controllers with convex optimization problems, and tasks that focus on the computational aspects of these optimization problems. The latter includes techniques to lower significantly the computational burden by solving a sequence of modest sized optimization problems, instead of a single problem with a large dimension. Tasks concerning the integration of the proposed scheduling technique with the traditional anti-windup approach are included also. The main motivation has been developing safe and reliable high performance control laws for protection of buildings, bridges, and similar large systems in civil infrastructure. Given the very large sizes, and the abrupt and destructive nature of the disturbances (e.g., earthquakes or wind gusts), there is a strong need for controllers that can adjust (or in the terminology of the proposed research, schedule) the control law to maximize the effectiveness of the inherently limited actuation mechanisms. The proposed approach can tolerate a great deal of uncertainty regarding the nature of the disturbance. It also guarantees overall stability and improved performance, while providing explicit tools for cost/benefit analysis regarding the size of the actuator vs. performance guarantees. The proposed research in computational aspects, in which the main problem is broken into a sequence smaller problems, is essential for application to large structures with very large models. Additionally, with some modifications, the proposed technique can be applied to structures equipped with several classes of semi-active devices. These aspects are necessary for the eventual acceptance of the proposed control framework by the structural community.