Comments on Dissipative Designs and the Tora Benchmark

Abstract

Building on physically motivated energy functions, dissipative designs often lead to relatively simple and intuitive controllers, along with guarantees for parameter-independent stability. We use the TORA example to illustrate some qualitative and quantitative issues concerning dissipative control, including the impact on performance of both lossless and dissipative design parameters. As happens in other non-linear systems, the analysis reveals substantial differences in performance-based design parameters selections, in different parts of the state space. The distinction between “large deviations” and “small deviations” dynamics suggests a controller structure that ignores energy dissipation of the actuator state, during large plant-state deviations, and moves to complete-state dissipation, in the small deviations mode. This alternative compensator provides for faster convergence with considerably lower actuation, than a single structure dissipative controller. Dissipative, observer based design is provided for the partial state feedback case, offering, much shorter time constants than static feedback compensators.