GelSight: High-Resolution Robot Tactile Sensors for Estimating Geometry and Force

Area of Science:

• Robotics
• Sensor Technology
• Computer Vision

Background:

• Existing tactile sensing technologies for robots have limitations.
• Key questions remain regarding the type and application of tactile information for robots.
• A soft sensor surface and high-resolution geometric sensing are crucial for competent tactile sensors.

Purpose of the Study:

• To discuss the development of GelSight, a vision-based optical tactile sensor.
• To emphasize the sensing principle and sensor design of GelSight for robotic applications.
• To review hardware and software supporting GelSight's use on robot hands.

Main Methods:

• Developed a vision-based optical tactile sensor (GelSight) with a soft elastomer contact surface.
• Designed an optical system and algorithms for measuring deformation, inferring shape, force, and slip.
• Fabricated different sensor versions and conducted experimental evaluations.

Main Results:

• GelSight measures geometry with very high spatial resolution, capturing object shape and surface tension.
• Contact force and slip can be inferred from sensor deformation.
• Experimental evaluations confirmed GelSight's performance in geometry and force measurement.

Conclusions:

• GelSight provides high-resolution measurement of shape and contact force.
• The sensor has successfully assisted robotic tasks like material recognition and in-hand localization.
• GelSight represents a significant advancement in tactile sensing for robotic manipulation.