Active Flow Control Using Adaptive Model Predictive Control with Online, Closed-Loop System Identification

Flow control is challenging for model-based and deep-learning control methods due to the complex, nonlinear dynamics of fluids. The present paper addresses this challenge by using predictive cost adaptive control (PCAC), which is an indirect adaptive control technique. PCAC performs continual, online, closed-loop linear system identification to facilitate output-feedback model predictive control. Operating from cold start with no prior modeling of the system dynamics, PCAC adapts automatically to changing flow conditions. This work assesses the performance of PCAC for several 2D flow-control scenarios. First, PCAC is used to control the velocity within a viscous boundary layer; next, PCAC is used to suppress turbulence. Finally, PCAC is used to specify the flow-separation location in order to delay stall on a high-angle-of-attack airfoil. The results demonstrate the applicability of PCAC to flow-control problems without prior control-oriented modeling or offline controller training.