Statistical Mechanics for Learning Algorithmic-Based Controllers: The Role or Physics in New Computational Models for Real-Time Control
Abstract
The objective of this research is to develop the foundations for real-time, feedback control algorithms, where the term ÒalgorithmÓ refers to a control action that is the outcome of an iterative sequence of mathematical steps, the result of which cannot be computed a priori in closed-form. To approach the objective the PIs will develop compressive tensor representations for partial differential equations, unify stochastic control in infinite dimensional spaces with non-parametric regression techniques with path integral control methods, and utilize geometric wavelets to create a hierarchy of dynamical models that can be used to generate a hierarchy of feedforward motion primitives or for designing local feedback controllers.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 27, 2017
- Source ID
- W911NF1610390
Entities
People
- Panagiotis Tsiotras
Organizations
- Army Contracting Command
- Georgia Tech Research Corporation
- United States Army