Exploiting Parallel Computing to Control Uncertain Nonlinear Systems in Real-Time

Control is a critical element in many applications and research such as experimental testing in real-time. Linear approaches for control and estimation have been widely applied to real-time hybrid simulation (RTHS) techniques in tracking the physical domain (plant). However, nonlinearities and highly uncertainties of the plant impose challenges that must be properly addressed using nonlinear control procedures. In this study, a controller is developed for such an uncertain nonlinear system by integrating a robust control approach with a nonlinear Bayesian estimator. A sliding mode control methodology synthesizes the nonlinear control law to provide stability and accurate tracking performance, and a particle filter algorithm estimates the full state of the plant using measured signals such as displacement. The Hybrid Simulation Management (HSM) code is developed to implement dynamic systems and the improved nonlinear robust controller. The HSM is integrated in a novel run-time substrate named CyberMech, which is a platform developed to enhance the performance of real-time cyber-physical experiments that supports parallel execution. A set of experiments with a highly uncertain nonlinear dynamic system demonstrates that the combination of advanced control techniques and high performance computation enhances the quality of real-time experimentation and potentially expands RTHS techniques capabilities.