An event-based opto-tactile skin

Area of Science:

• Robotics and Artificial Intelligence
• Sensor Technology
• Neuromorphic Engineering

Background:

• Traditional tactile sensors often rely on repetitive scanning, limiting responsiveness and efficiency.
• Developing large-area, flexible tactile sensing remains a challenge for soft robotics and human-computer interaction.

Purpose of the Study:

• To present a neuromorphic, event-driven tactile sensing system for soft, large-area skin.
• To evaluate the system's press localization accuracy and performance under data reduction.
• To explore the potential for low power consumption and computational load.

Main Methods:

• Integration of Dynamic Vision Sensors (DVS) with a flexible silicone optical waveguide skin.
• Utilizing a stereo vision setup for triangulation-based press localization.
• Employing Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for event data processing.

Main Results:

• Achieved a Root-Mean-Squared Error (RMSE) of 4.66 mm for press localization over a 4,620 mm² area.
• Demonstrated functional press localization even when event data was reduced to 1/1,024, with a slight increase in error to 9.33 mm.
• Reported a detection latency with a characteristic width of 31 ms.

Conclusions:

• The neuromorphic, event-driven tactile sensing system shows significant potential for wide-area, flexible, and responsive tactile sensing.
• The system's robustness to sparse event data highlights its suitability for low-power and computationally efficient applications.
• This approach offers a promising direction for advanced tactile sensing in soft robotics and interactive environments.