Data-driven control of oscillator networks with population-level measurement

Area of Science:

• Complex systems
• Nonlinear dynamics
• Network science

Background:

• Controlling complex networks of nonlinear limit-cycle oscillators is crucial for engineering and natural sciences.
• Existing methods often rely on detailed biophysical or simplified phase models.
• Data-driven control without prior model assumptions is a less developed research area.

Purpose of the Study:

• To develop an online learning control strategy for oscillator networks directly from data.
• To address the challenge of controlling systems without pre-existing models or extensive datasets.
• To enable effective regulation of synchrony in diverse oscillator network configurations.

Main Methods:

• Leveraging current network dynamics for iterative online control learning.
• Employing a single input and a noisy population-level output measurement.
• Utilizing numerical simulations for technique validation.

Main Results:

• The proposed technique effectively regulates synchrony in various oscillator networks.
• Successful control is achieved after a small number of trials.
• The approach demonstrates robustness to system variations.

Conclusions:

• The developed online learning method offers a powerful data-driven approach to control complex oscillator networks.
• This technique bypasses the need for global system models, simplifying control implementation.
• The method shows promise for applications requiring adaptive and efficient network control.