SimTac: A Physics-Based Simulator for Vision-Based Tactile Sensing with Biomorphic Structures

Area of Science:

• Robotics and Artificial Intelligence
• Bio-inspired Engineering
• Materials Science

Background:

• Biological tactile sensing relies on complex morphologies (e.g., fingers, trunks) for rich interactions.
• Robotic tactile sensors are typically limited to simple planar designs, neglecting biomorphic potential.

Purpose of the Study:

• To introduce SimTac, a physics-based simulation framework for designing and validating biomorphic tactile sensors.
• To bridge the gap between bio-inspired design and practical robotic tactile sensing.

Main Methods:

• Utilized particle-based deformation modeling for accurate physical simulation.
• Employed light-field rendering for photorealistic tactile image generation.
• Integrated a neural network for predicting mechanical responses across diverse geometries and materials.

Main Results:

• Successfully designed and validated physical sensor prototypes inspired by biological tactile structures.
• Demonstrated SimTac's effectiveness in Sim2Real tasks: object classification, slip detection, and safety assessment.
• Showcased accurate and efficient simulation capabilities for various geometries and materials.

Conclusions:

• SimTac expands the design space for tactile sensors by integrating morphology and sensing.
• The framework facilitates the development of robust tactile sensing systems for unstructured environments.
• Enables bio-inspired tactile sensors that enhance robotic interaction capabilities.