Joint Sensor Node Selection and State Estimation for Nonlinear Networks and Systems

TL;DR

This paper introduces a new methodology for sensor node selection and state estimation in nonlinear networks, integrating the two problems into a unified approach and demonstrating its effectiveness on various nonlinear systems.

Contribution

It presents a novel integrated approach for sensor selection and state estimation in nonlinear networks, addressing limitations of existing graph-based methods and separate problem treatments.

Findings

Effective sensor selection and state estimation demonstrated on nonlinear systems.

The approach outperforms traditional methods in numerical tests.

Codes are publicly available for further research.

Abstract

State estimation and sensor selection problems for nonlinear networks and systems are ubiquitous problems that are important for the control, monitoring, analysis, and prediction of a large number of engineered and physical systems. Sensor selection problems are extensively studied for linear networks. However, less attention has been dedicated to networks with nonlinear dynamics. Furthermore, widely used sensor selection methods relying on structural (graph-based) observability approaches might produce far from optimal results when applied to nonlinear network dynamics. In addition, state estimation and sensor selection problems are often treated separately, and this might decrease the overall estimation performance. To address these challenges, we develop a novel methodology for selecting sensor nodes for networks with nonlinear dynamics. Our main idea is to incorporate the sensor selection problem into an initial state estimation problem. The resulting mixed-integer nonlinear optimization problem is approximately solved using three methods. The good numerical performance of our approach is demonstrated by testing the algorithms on prototypical Duffing oscillator, associative memory, and chemical reaction networks. The developed codes are available online.