Data-Driven Adaptive Control of Shape Evolution with Regime Changes

Abstract

The main goal of this project is to enable data-driven adaptive control of shapes evolving dynamically, where the input-output dynamics abruptly change between regimes of the input domain. Toward this goal, we investigate multiple dynamic state space models for shape evolution and develop the associated control modeling capable of handling regime changes. We also investigate faster rate learning methods for fitting these nonstationary dynamics based on the active machine learning principle. Finally, we will integrate the new methodological developments with an approximate algorithm for the Markov decision process to demonstrate the new modeling and control algorithm interplay. The proposed research efforts are well aligned with the research themes of AFOSR s Dynamical Systems and Control Theory Program. The efforts contribute to fundamental research in dynamic control modeling in shape spaces. The proposed efforts also contribute to nonlinear control estimation and novel computational techniques associated with the new control modeling. In the statistical shape literature, the space of shapes is represented as a non-Euclidean space, often a Riemannian manifold. Our efforts will focus on state space modeling with continuous states in a Riemannian manifold. We also invent new non-parametric control modeling of the regime changes with Euclidean inputs and non-Euclidean outputs. Control modeling associated with manifold state spaces and regime changes is a novel research topic that has not been addressed before. The fundamental research would broaden the application of data-driven adaptive control to complex dynamical processes possessing unique characteristics- shape state space and regime changes. The applications are directly relevant to future U.S. Air Force and Space Force Systems. For instance, we will demonstrate the applicability of the proposed control modeling for dynamic, autonomous calibration of additive manufacturing in collaboration with the Air Force Research Lab, Manufacturing and Materials Directorate. The application example would enable enhanced capabilities in precise fabrications of air force materials through additive manufacturing and 3D printing for future U.S. Air Force and Space Force systems. In addition, the proposed research will benefit other air force research, e.g., dynamic control of nanomaterial self-assemblies (also additive material processes). Another board application is control for AF materials produced through additive or traditional deductive processes.

Document Details

Document Type
DoD Grant Award
Publication Date
Mar 14, 2024
Source ID
FA95502310673

Entities

People

  • Anuj Srivastava

Organizations

  • Air Force Office of Scientific Research
  • Florida State University
  • United States Air Force

Tags

Readers

  • Computational Fluid Dynamics (CFD)
  • Control Systems Engineering.
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Machine Learning Algorithms
  • Space
  • Space - Spacecraft Maneuvers