Data-Driven Controller and Multi-Gradient Search Algorithm for Morphing Configurations

n this paper, we investigate the use of morphing configurations in scenarios where safety, optimality, and stability are important aspects of the operation. We develop an online, data- based framework for the airfoil to be able to optimally morph the airfoil thickness. Considering an initial shape, we aim to search online a shape with a reduced value of the drag coefficient, an increased value of the lift coefficient, and a “reasonable” value of the pitching moment coefficient that obeys the geometric constraints and guarantee a safe transition from an initial shape to the final shape. We then use a surrogate model, based on deep neural networks combined with a multi-gradient search algorithm to obtain a list of shapes with a higher value of cl/cd, an increased value of cl, a reduced value of cd, and a desired reduction of cm. Finally, we use a data-driven shape controller to guarantee a safe transition from the initial shape to the final shape while following a smooth trajectory. Experimental numerical results show the efficacy of the proposed solution for different trajectories.