Efficient Sensor Node Selection for Observability Gramian Optimization.

Area of Science:

• Control Systems Engineering
• Optimization Theory
• Signal Processing

Background:

• Identifying critical sensor nodes is crucial for monitoring large-scale dynamical systems.
• Existing sensor selection methods may face challenges with computational complexity and robustness in real-world applications.

Purpose of the Study:

• To develop and evaluate novel algorithms for sensitive sensor node selection in linear time-invariant discrete-time dynamical systems.
• To compare the performance of new methods against existing techniques using the observability Gramian determinant.

Main Methods:

• Proposed two novel algorithms: an approximate convex relaxation with Newton's method and a gradient greedy method.
• Utilized the matrix determinant of the observability Gramian for sensor subset evaluation.
• Derived the gradient and Hessian for the proposed optimization methods.

Main Results:

• The approximate greedy method demonstrated superior runtime efficiency for systems where sensor count approximates latent system dimensions.
• Semidefinite programming (SDP) based convex relaxation proved effective for high-dimensional, randomly generated systems.
• Pure greedy selection exhibited the most stable optimization results, particularly with real-world datasets, despite some degradation in other methods.

Conclusions:

• Novel sensor selection algorithms offer improved performance trade-offs in terms of speed and accuracy.
• The choice of sensor selection method should consider system characteristics, dimensionality, and data type (numerical vs. real-world).
• Further research may focus on enhancing the robustness of convex relaxation methods for practical applications.