Time-varying actuation and structure modification for assessing and enhancing network resilience
Abstract
Despite important advances in designing distributed coordination, cooperation, and decision-making algorithms, multi-agent networks have proven fragile to targeted attacks, as local and well-orchestrated actions have rapidly cascaded into network-wide destructive perturbations. This has been demonstrated both in academic research and by real-world events. Given the paramount importance that resilience plays in the autonomous operation of network systems, novel theories, methodologies, and tools are critically needed. We need notions that characterize network resilience and allow us to understand strengths and vulnerabilities against adversaries. Obtaining such characterizations is difficult because resilience and performance are complex functions of the network s and adversary s capabilities, knowledge, and resources, the interconnection structure, and the physical limitations on remedial and adversarial actions. At the same time, we also need novel design methodologies that protect multi-agent networks and adaptively manage their interconnection over time to achieve performance guarantees in military missions with a high degree of uncertainty, complexity, and time urgency. The goal of this project it to assess and control resiliency of networks by providing ways to measure it, identify vulnerabilities and areas of improvement, and leverage time-varying actuation and interconnection to develop novel resiliency-enhancing control methodologies. This project seeks to address challenges posed by a diversity of scenarios affecting the secure operation of multi-agent networks against strategic adversaries, such as the availability of information to individual agents and to the network designer, offline versus online operation, the impact of changes in the network state and the environment, and tradeoffs of centralized versus distributed computation. We are particularly interested in developing metrics, tools, and methods to enhance resilience that account for scenarios with partial, local, or incomplete knowledge and leverage time-varying actuation and network topologies. Understanding and taking advantage of this requires measures of resiliency for both static and dynamic networks and efficient ways to compute them. The expected outcomes of the project are understanding of fundamental trade-offs between resiliency and attack/defense capabilities in multi-agent networks; the identification of structural principles that help make network resilient in linear and nonlinear scenarios; and design guidelines and control strategies for attack detection and network adaptation that take advantage of time-varying actuation and interconnection. An increased understanding of the key factors that determine network resilience promises to be of significant impact in a wide range of scenarios. Prior to mission deployment, by providing critical insights and guidance as to how to structure network interactions and how to react to specific types of attacks. During missing deployment, by providing novel distributed computational procedures to re-assess resilience and endowing networks with adaptive control methodologies that take advantage of time-varying actuation of agent nodes, clusters, and interconnections.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 28, 2023
- Source ID
- W911NF2310138
Entities
People
- Jorge Cortés
Organizations
- Army Contracting Command
- United States Army
- University of California, San Diego