Specification, Synthesis, and Verification of Software-based Control Protocols for Fault-Tolerant Space Systems
Abstract
In this one-year project, we focused on the control and learning for systems under stochastic uncertainties. We report two sets of results. The first one is motivated by correct-by-construction synthesis for systems with uncertainty in the state due to partial and/or noisy measurements. We developed a new finite-state abstraction technique for such systems. This particular problem was motivated by planning for autonomous space operations. The second one focuses on control and learning in systems in which there is an embedded data-classifier that imperfectly (characterized as stochastically) generates labels from a finite set. Our main contribution was showing how inference techniques for discrete Markov random fields can be applied to learning and control tasks which depend on the output of a noisy classification process.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 16, 2016
- Accession Number
- AD1020450
Entities
People
- Ufuk Topcu
Organizations
- University of Texas at Austin