Feedback-based Deep Learning for Uncertain Dynamical Systems
Abstract
The University of Florida is requesting a grant in the amount of $872,119.00 (base period: $283,377.00; Year One (Option): $290,621.00; Year 2 (Option): $298,120.00) to develop robust and adaptive real-time Lyapunov-based deep learning algorithms, encoded in software, to yield assuredly robust, safe, and optimal behaviors for a general class of agile (e.g., performance envelopes with nonlinear dynamical effects) NCA agents that are applicable and scalable across a network with intermittent feedback and communications. Specifically, this project will focus on five specific aims through a one-year base period followed by two optional periods. During the base period, Robust Adaptive Lyapunov-based Deep Neural Networks (RALb-DNN) algorithms (Aim 1) will be developed for NCA agents motivated by the need to develop real-time adaptation that is robust and verifiable with assurances derived from a nonlinear stability analysis.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 12, 2024
- Source ID
- FA86512410018
Entities
People
- Warren E Dixon
Organizations
- Air Force Research Laboratory
- United States Air Force
- University of Florida