Compositional and Hierarchical Design of Network Control Systems

Abstract

This project will develop a compositional and hierarchical framework to manage thecomplexity of network control systems. The aim in hierarchical control is to compose acontrol action for an aggregate model and to rene this action to lower-level controllers.The aim in compositional design is to expose the system as an interconnection of subsystemsand to derive guarantees for each subsystem under appropriate assumptions on theothers. Taken together these approaches break apart an intractably large problem intosubproblems of manageable size. In prior work the PI developed a compositional approachto derive stability, performance, and safety guarantees. This project will addressa richer class of specifications, including those expressed in temporal logic, that surpassthe traditional objectives in control theory. Examples of properties expressed in temporallogic include fairness (whenever some condition occurs, another condition is guaranteedto occur), repeated reachability (a certain condition occurs innitely often), andsequentiality (a condition occurs only after another). Although formal methods fromcomputer science are well suited to such specications, they do not scale to continuousstatemodels with more than a few state variables. The proposed compositional andhierarchical framework aims to overcome this limitation and enable applications to networkcontrol systems of ever increasing complexity. The project will further improvethe scalability of formal synthesis and verification engines by developing reachabilitymethods that exploit sparsity patterns and symmetries in the system model. Buildingcorrect-by-design systems that guarantee high-level specifications is a critical ability forAir Force systems operating in uncertain and adversarial environments, including autonomoussystems where progress has been held back by the lack of scalable validationand verification methods.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 11, 2018
Source ID
FA95501810253

Entities

People

  • Murat Arcak

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of California Regents

Tags

Fields of Study

  • Computer science

Readers

  • Data Mining and Knowledge Discovery.
  • Distributed Systems and Data Platform Development
  • Mathematical Modeling and Probability Theory.