Data-driven modeling and identification of a bistable soft-robot element based on dielectric elastomer.

Area of Science:

• Robotics
• Materials Science
• Mechanical Engineering

Background:

• Dielectric elastomer (DE) actuators are crucial for soft robotics, offering unique electromechanical properties.
• Bistable soft robotic systems require precise modeling to achieve controlled switching between states.
• Existing models often struggle with nonlinearities and unmodeled effects inherent in soft actuators.

Purpose of the Study:

• To develop and experimentally validate a hybrid modeling framework for bistable soft robotic systems.
• To integrate physics-based analytical modeling with data-driven approaches for enhanced accuracy.
• To capture the complex nonlinear and dynamic behaviors of dielectric elastomer-driven soft robots.

Main Methods:

• A hybrid modeling framework combining physics-based analytical models and radial basis function (RBF) networks.
• Derivation of a physics-based model for electromechanical coupling and dynamic behavior of DE actuators.
• Augmentation of the analytical model with RBF networks trained on experimental data to address discrepancies.

Main Results:

• Successful development of a hybrid model capable of capturing the nonlinear dynamics of bistable soft robots.
• Experimental validation demonstrating the framework's accuracy in predicting system behavior.
• Improved modeling accuracy by incorporating data-driven RBF networks to account for unmodeled effects.

Conclusions:

• The hybrid modeling framework provides a robust approach for characterizing bistable soft robotic systems.
• This integrated method enhances the predictive capabilities for dielectric elastomer-driven soft robots.
• The validated framework facilitates the design and control of advanced soft robotic applications.