We are aiming at developing novel machine learning and artificial intelligence algorithms, with application emphases on aircraft design optimization, unmanned aerial vehicles (UAVs), etc.
Aircraft Design Optimization through Machine Learning Surrogates
We developed reduced-dimensional deep neural network surrogate models for directly predicting optimal designs, which realize aircraft design optimization in a fast interactive manner.
B-Spline-based Generative Adversarial Networks
We developed the B-spline-based generative adversarial networks (BSplineGAN) to generate realistic smooth airfoil shapes (compared with direct B-spline parameterization) for intelligent parameterization and automatic dimensionality reduction.
Artificial Intelligence-based Trajectory Design for UAVs
We predicted optimal trajectory for electric drones using long short-term memory (LSTM) networks with ~99.8% generalization accuracy, which outperformed multiple-output Gaussian Processes (MOGP).