Model-based data-driven learning methods for Optimal Feedback Control
Abstract
This proposal addresses a critical challenge in control design and applications, namely efficient computational methods for optimal feedback control of high-dimensional nonlinear systems, which has been a main bottleneck limiting implementations of many nonlinear feedback control methodologies on real world applications. We plan to accomplish this goal by designing a modelbased data-driven learning method for high-dimensional Hamilton-Jacobi-Bellman (HJB) equations. The essential idea of the proposed research is to integrate physical models of control systems with causality-free type of computational methods to generate information rich data sets, from which neural network can be trained as an efficient high-dimensional approximation tool to learn the solution to HJB equations and the associated optimal feedback control.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 21, 2022
- Source ID
- FA95502110113XX0
Entities
People
- Qi Gong
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of California, Santa Cruz