State observation and sensor selection for nonlinear networks

Area of Science:

• Complex Systems
• Nonlinear Dynamics
• Systems Biology

Background:

• Many natural and engineered systems are nonlinear networks.
• Predicting and controlling these systems requires knowing their full state.
• Network states are often unknown, with only partial measurements available.

Purpose of the Study:

• To develop a general optimization-based approach for observing nonlinear network states.
• To identify optimal subsets of variables for observation.
• To reveal fundamental limitations and trade-offs in network observability.

Main Methods:

• Formulated an optimization-based approach for state observation.
• Investigated the impact of measurement fraction and observation length on estimation error.
• Analyzed the limitations of graph-theoretic methods in nonlinear systems.

Main Results:

• Identified fundamental trade-offs between observation parameters and estimation error.
• Demonstrated that graph-theoretic approaches are insufficient for practical state estimation in nonlinear systems.
• Successfully identified key components in biological and combustion networks for full state determination.

Conclusions:

• The proposed optimization method effectively reconstructs nonlinear network states from limited data.
• System dynamics are critical for observability, challenging purely structural analyses.
• Results pave the way for novel sensing strategies to improve prediction and control of complex systems.