Dear All:
We are pleased to announce the following Fall 2023 online seminar at the Forum on Robotics & Control Engineering (FoRCE, https://www.youtube.com/playlist?list=PLW4eqbV8qk8Z3j1YqD8jCJExA-aiMGFZb):
Title: Physics Informed Neural Networks and Theory of Functional Connections: Applications to Optimal Control
Speaker: Dr. Roberto Furfaro (Professor and Director of the Space Systems Engineering Laboratory at the University of Arizona)
Abstract: Since the first paper from Raissi, Perdikaris & Karniadakis (2019, Journal of Computational Physics), there has been a lot of interest in pursuing the development and applications of Physics-Informed Neural Networks (PINN) to solving ODEs and PDEs. Concurrently, in the aerospace world, there has been an increasing need for computational methods that enable autonomy and real-time decision making, calling for the development of new frameworks for the numerical solution of optimal control problems. Here, we discuss the development and implementation of a framework that merges PINN and the Theory of Functional Connections (TFC) to accurately, fastly and efficiently solve optimal control problems for aerospace guidance and control. TFC has been developed to generate a class of functions that analytically satisfy specified boundary conditions by combining a boundary-free function and problem-specific constrained expressions. PINNs have been recently developed to enforce differential equations arising from the application of physical first-principles to the training of neural networks. The combination of TFC and PINN enables solving ODEs & PDEs by training NNs in a data-driven fashion to minimize the residual in the equations domain while analytically satisfying the BCs for highly accurate solutions. In this talk, we will review the development of the TFC-PINN framework to train shallow and deep NNs that approximate open-loop and closed-loop controllers for a large class of optimal control problems. We will demonstrate how such framework can be applied to a variety of problems that arise in the aerospace world and we discuss the potential to employ NNs for real-time guidance and control of aerospace systems.
Date: December 1, 2023, 12:00 Eastern Time
Platform: MS Teams – Login instructions are provided at the bottom of this post!
Tansel Yucelen, University of South Florida, yucelen@usf.edu
K. Merve Dogan, Embry-Riddle Aeronautical University, dogank@erau.edu
Microsoft Teams meeting
https://teams.microsoft.com/l/meetup-join/19%3ameeting_Mzc3ZGJiZmMtNDRkYi00MzIxLWFkYTEtODljOTg5YWQzYzZj%40thread.v2/0?context=%7b%22Tid%22%3a%22741bf7de-e2e5-46df-8d67-82607df9deaa%22%2c%22Oid%22%3a%224c398941-082c-4b5f-bd3c-503527bb01c2%22%7d
Meeting ID: 281 389 576 334
Passcode: oD3Xm7
