Quantifying network controllability and observability using optimal control and estimation metrics
Abstract
Success in future battle spaces will require a foundational understanding of complex multi-agent dynamical networks in order to tightly coordinate distributed technological assets and human teams in dynamic, uncertain, and adversarial environments. However, traditional analysis and design tools from systems and control theory are not sufficient to handle the emerging complexity. There is a profound need to explore how modern computation and communication resources can be utilized to address these challenges. The proposed research will: (1) quantify and visualize network controllability and observability using metrics based on fundamental optimal feedback control and state estimation problems; (2) develop algorithms for design of optimal sensor and actuator topologies and feedback information structures based on these metrics; and (3) illustrate the theoretical and methodological results with comprehensive case studies in various application domains. Thrust 1: Quantifying and visualizing network controllability and observability. There are many ways to quantify how easy or difficult is it to control and observe the state of a dynamical network with available actuators and sensors. Prominent recent work has focused on binary and open-loop metrics that can give limited and misleading views of network controllability and observability. To incorporate essential notions of closed-loop feedback, robustness, and information structures, we will utilize quantitative metrics for network controllability and observability based on several optimal control and estimation problems. The associated optimal cost functions can be used as maps for visualizing controllability and observability, and we will study these maps for various classes of random and regular networks to build intuition. We will also explore the use of distributed algorithms and distributed computing platforms to scale computations to large networks. Thrust 2: Algorithms for network topology and feedback information structure design. We will utilize the metrics from Thrust 1 to develop algorithms for optimizing sensor and actuator topologies and feedback information structures to enhance network controllability and observability properties. This can be accomplished by adding sensors and actuators at locations in the network where they provide the most performance and robustness and by adding the most effective communication links between sensors and actuators. We will analyze and compare the theoretical and empirical properties of two classes of methods based on combinatorial greedy algorithms and convex relaxation algorithms. We will also investigate and elucidate how sensor and actuator topologies and information structures impose fundamental limits of feedback control and estimation in large dynamical networks. Thrust 3: Application area case studies. The results in Thrusts 1 and 2 will be applied to three important application domains. In particular, we will study intelligence, surveillance, and reconnaissance scenarios involving autonomous mobile robot teams. We will also study neuronal brain networks, in which our tools can elucidate how new technologies could be used to understand and control dynamical processes in the brain based on fundamental principles of feedback and robustness. Finally, we will study placement of energy storage devices and phasor measurements units for frequency and rotor angle control in future power networks.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 24, 2018
- Source ID
- W911NF1710058
Entities
People
- Tyler Summers
Organizations
- Army Contracting Command
- United States Army
- University of Texas at Dallas