Session: DAC-01-01-Control Co-Design

Paper Number: 89957

89957 - A Constraint-Handling Technique for Parametric Optimization and Control Co-Design 

For designing dynamic systems, control co-design (CCD) helps engineers integrate the design processes of plant and control systems. Stability and control saturation are two examples of constraints needed to be considered in the design problem. Sometimes, the capability of controllers or actuators may change over the stage of development due to cost, availability, or other factors. Therefore, a parametric study of optimal designs with respect to different control limits of actuation should be analyzed since the value is undetermined or uncontrollable. However, repeatedly solving optimization problems by adjusting the parameter value may be inefficient. Parametric optimization can accelerate the process of finding optimal solutions with various parameter values, but there lack techniques for handling nonlinear constraints. This paper proposes a novel constraint-handling technique for parametric optimization and CCD problems with constraints. Using a machine-learning classification approach called support vector machine (SVM), constraint boundaries are approximated and calibrated during optimization. A test problem and a case study of an elastic inverted pendulum on a cart are used as benchmarking problems. Performance and computational cost are compared to the existing parametric optimization technique, Predicted Parameterized Pareto Genetic Algorithm (P3GA). Results show the P3GA with the proposed method outperforms the original P3GA in both qualitative and quantitative perspectives.

Presenting Author: Ying-Kuan Tsai Texas A&M University

Presenting Author Biography: I am a Ph.D. student at Texas A&M University.

Authors:

Ying-Kuan Tsai Texas A&M University
Richard Malak Jr. Texas A&M University