Dr. Ivan Tkachenko
Samara University, Russia
Speech Title: Methodological Approaches to the Creation and Functioning of Serial Robotic Production of Small Spacecrafts
Abstract: Embedding the intrinsic symmetry of a flow system in training its machine learning algorithms has become a significant trend in the recent surge of their application in fluid mechanics. This work leverages the geometric symmetry of a four-roll mill (FRM) to enhance its training efficiency. Stabilizing and precisely controlling droplet trajectories in a FRM is challenging due to the unstable nature of the extensional flow with a saddle point. Extending the work of Vona & Lauga (2021), this study applies Deep Reinforcement Learning (DRL) to effectively guide a displaced droplet to the center of the FRM. Through direct numerical simulations, we explore the applicability of DRL in controlling FRM flow with moderate inertial effects, i.e., Reynolds number~O(1), a nonlinear regime previously unexplored. The FRM's geometric symmetry allows control policies trained in one of the eight sub-quadrants to be extended to the entire domain, reducing training costs. Our results indicate that the DRL-based control method can successfully guide a displaced droplet to the target center with robust performance across various starting positions, even from substantially far distances. This study presents new advances in controlling droplet trajectories in more nonlinear and complex situations, with potential applications to other nonlinear flows, such as polymeric fluids with elastic nonlinearity. The geometric symmetry used in this cutting-edge reinforcement learning approach can also be applied to other control methods.
Biodata: Ivan Tkachenko, Deputy Rector, Director of the Institute of Aviation and Rocket and Space Technology of Samara University, Ph.D., Associate Professor. A specialist in the development of small spacecraft and their constellations. The head of the project for the creation of the university small satellites' constellation of the "AIST" series. Currently, a project is being implemented to develop technologies for serial robotic satellite production, the results of which will be implemented at Roscosmos enterprises.
Prof.
Lei Lei
Nanjing University of Aeronautics and Astronautics, China
Speech Title: Research on Intelligent Cooperation Theory of Unmanned Aerial Vehicular Swarms Based on Digital Twins
Abstract: Lei Lei received the B.S. degree in electronic and information engineering from Northwestern Polytechnical University, Xi’an, China in 2002, and the Ph.D. degree in communication and information systems from Nanjing University of Aeronautics and Astronautics, Nanjing, China in 2008. He is currently a professor and doctoral supervisor with the College of Electronic and Information Engineering, Nanjing University of Aeronautics and Astronautics. His current research directions include intelligent networking and collaboration technology of aircrafts and air-space information system digital twin technology. In recent years, he has undertaken over 40 projects of various types. He won the first and third prizes for National Defense Science and Technology Progress, the second and third prizes for National Defense Technology Invention. In recent years, he has published more than 50 papers in international mainstream journals and conferences in communication networks fields such as IEEE NETWORK, IEEE IoT-J, IEEE TVT, IEEE CL.
Biodata: Unmanned aerial vehicle (UAV) swarm networking and collaboration have significant prospects in both civilian and military applications, due to its remarkable properties in cooperative efficiency, reduced risks, and operational cost. In recent years, breakthroughs in artificial intelligence (AI) technologies such as deep reinforcement learning have provided new methods for intelligent cooperative control of swarms. However, how to improve the training efficiency in a high-fidelity manner has become a key bottleneck in applying deep reinforcement learning to intelligent cooperative control of swarms. In order to break through this bottleneck, we propose a new approach to achieve intelligent cooperative control of swarms through multi-agent deep reinforcement learning based on building a digital twin (DT) system of swarms. Our contributions consist of the following aspects: (1) We proposed a DT based training method of deep reinforcement learning for the intelligent cooperation decision model of UAV swarms. Our method realizes the continuous evolution of the decision model; (2) We proposed a new method for cooperative trajectory planning based on digital twins and deep reinforcement learning to realize the dynamic optimization of the decision model under multiple constraints; (3) We proposed a linear fusion method for UAV detection information and establish a scalable deep reinforcement learning decision-making model for the cooperative electronic reconnaissance problem.
Prof.
Weitao
Wu
Nanjing University of Science and Technology,
China
Speech Title: Physics Informed Machine Learning in Fluid Flow and Heat Transfer
Abstract: Flow and heat transfer phenomena widely exist in industry and nature. Nowadays, researchers have been able to obtain high-precision physical fields by computational or experimental techniques, which helps the understanding of the mechanism or the guidance of the engineering design. Through decades of studies, although people have accumulated huge size of computational and experimental data, when working on a new problems even similar problem with different conditions, usually it is still necessary to re-simulate or re-experiment. In recent years, deep learning has demonstrated great ability on extracting features from thus accurately predicting physical fields, and prediction speed by deep learning is usually several orders of magnitude faster. This presentation aims to discuss the recent advances of our group in physics informed machine learning in fluid flow and heat transfer.
Biodata: Dr. Wei-Tao Wu is a professor in School of Mechanical Engineering at Nanjing University of Science and Technology. Dr. Wu received his Ph.D. degree from the Department of Mechanical Engineering of Carnegie Mellon University, USA, and his B.S. degree from Xi’an Jiaotong University, China. Dr. Wu’s research interests lie in multiphase flow, non-Newtonian fluid, aerodynamics and physics-informed machine learning. His work covers mathematical modeling, computational simulation and data mining. Within recent 5 years, Dr. Wu has authored or co-authored over 100 peer-reviewed journal papers, and he has also served as an editorial board member of Fluids and a member of Physics and Aerodynamics Committee of Chinese Aerodynamics Research Society.