Optomechanical reservoir computing

Area of Science:

• Physics
• Engineering
• Computer Science

Background:

• Nonlinear dynamics are essential for physical reservoir computing, enabling complex input-output mappings.
• Existing physical reservoirs often rely on a single source of nonlinearity, limiting their predictive capabilities.
• A gap exists between theoretical analysis and practical evaluation of reservoir computer performance.

Purpose of the Study:

• To introduce a novel optomechanical reservoir that integrates multiple sources of nonlinearity.
• To analyze and characterize the nonlinear frequency content generated by the reservoir.
• To advance the rational design of physical reservoir computers for embodied intelligence.

Main Methods:

• Developed an optomechanical reservoir combining nonlinear springs and nonlinear optical fiber sensing.
• Employed a novelty search on simulated reservoirs to identify high-performing designs.
• Validated promising designs through experimental testing.
• Introduced a frequency content metric to characterize reservoir nonlinearity.

Main Results:

• The combined nonlinear springs and optical fibers significantly expanded the reservoir's nonlinear frequency content.
• Novelty search identified and experimental validation confirmed high-performing optomechanical reservoir designs.
• The frequency content metric effectively characterized the reservoir's nonlinear capabilities.

Conclusions:

• The optomechanical reservoir, leveraging multiphysics nonlinearity, demonstrates enhanced performance for physical reservoir computing.
• The developed analytical techniques and multiphysics designs pave the way for more effective embodied intelligence.
• This work bridges the gap between fundamental reservoir capabilities and practical applications.