Fin-Bayes: A Multi-Objective Bayesian Optimization Framework for Soft Robotic Fingers.

Area of Science:

• Robotics
• Computational Design
• Materials Science

Background:

• Soft robotics offers high performance, flexibility, and safe interaction.
• Computational design is key but faces limitations in assessment fidelity, iteration count, and single-objective optimization.
• Bayesian approaches are efficient for expensive-to-analyze systems in design space exploration.

Purpose of the Study:

• To address shortcomings in computational design for soft robotics.
• To develop a framework for rapid, accurate characterization and extensive design space exploration.
• To optimize Fin Ray grippers for multiple objectives like compliance and contact force.

Main Methods:

• Utilized a nonlinear Finite Element Modeling suite for accurate characterization.
• Conducted extensive physical testing with an automated rig and 3D printed robotic fingers.
• Employed a multiobjective Bayesian optimizer to explore trade-offs between compliance and contact force.

Main Results:

• Achieved rapid and accurate characterization of robotic designs.
• Explored a significantly larger design space, discovering novel, high-performance designs.
• Identified promising Fin Ray gripper designs balancing compliance and contact force.

Conclusions:

• The developed computational framework enhances soft robotics design optimization.
• The optimized Fin Ray grippers are validated through 3D printing and practical application.
• This approach enables the discovery of superior robotic designs with improved performance characteristics.