Risk Analysis of Stochastic Nonlinear Dynamical Networks Via Lifting Operators

Abstract

In this project, we developed operator theoretic tools to study convergence of Carleman lifting of time-varying nonlinear systems. We quantified explicit error bounds for finite-section of the lifted system and proved that the truncation error converges exponentially to zero as the truncation length increases. We utilized our findings to study Hamilton-Jacobi-Bellman (HJB) equation and proposed a quadrization method to lift and represent the HJB equation as a semi-Riccati operator equation. An exact iterative algorithm to calculate the sub-blocks of the solution to the semi-Riccati operator equation up to an arbitrary precision was proposed. Furthermore, we developed a methodology to study and quantify systemic risk measures in various dynamical networks subject to Gaussian or stable non-Gaussian noise. More specifically, our findings show that for a class of nonlinear dynamical networks one can quantify the value of a family of risk measures using spectral mode decomposition techniques. Our theoretical results have been applied to various applications including platoon of autonomous vehicles, swarm of multiple robots, rendezvous and synchronization in multi-agent systems, and power networks.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 11, 2022
Accession Number
AD1230408

Entities

People

  • Nader Motee

Organizations

  • Lehigh University

Tags

Readers

  • Calculus or Mathematical Analysis
  • Mathematical Modeling and Probability Theory.
  • Unmanned Aerial System (UAS) Autonomous Capabilities and Mission Reconnaissance.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms
  • Autonomy
  • Autonomy - Autonomous System Control