Robots Get All Touchy-Feely

Replicating the remarkable tactile sensing abilities of the human hand has thus far remained an elusive goal in the field of robotics. But significant progress has been made in recent years. One particularly promising area of development has been in the creation of GelSight sensors. These sensors employ a combination of soft elastomeric materials and high-resolution imaging to detect and interpret surface textures and forces in a manner similar to human touch. The system consists of a rubbery material coated with a reflective surface, creating a deformable gel pad that acts as a sensitive tactile interface.

When a GelSight sensor comes into contact with an object, the gel pad compresses and conforms to the object's shape and surface texture. The deformation of the gel is then captured using high-resolution cameras, which record the changes in the reflective surface. These camera images are processed using sophisticated algorithms to extract detailed information about the object's surface, such as its texture, roughness, and even subtle changes in shape.

These sensors do have their limitations, however, that make them unsuitable for some use cases. For example, when the elastomer is uniformly deformed, or when it reaches its maximal point of deformation, it will produce inaccurate force measurements. To get to a human-like level of tactile sensing, improvements to this platform are still needed.

One such improvement was recently described by researchers from Queen Mary University in London. They have developed a lightweight and low cost sensor based on the GelSight design, but with some important enhancements that allow it to produce more accurate force measurements than existing systems, even when the elastomer is stretched to its limits. Called L3 F-TOUCH, these fingertip sensors produce high-resolution data and can transmit it to a processing unit wirelessly.

Rather than directly measuring surface deformations of the gel, L3 F-TOUCH instead incorporates an integrated mechanical suspension structure with an ARTag attached to its base. As the elastomer is displaced, this ARTag moves and can be tracked by the camera. An algorithm can translate these movements into force contacts along the x, y, and z axes of the surface of the sensor. This information is directed, via a mirror, to the same camera that images the surface of the gel to detect tactile information. That setup keeps the device lightweight and low cost.

A series of experiments confirmed that the L3 F-TOUCH sensor could indeed overcome the limitations of traditional GelSight sensors. The researchers also demonstrated that their system achieved higher levels of measurement accuracy, and that the L3 F-TOUCH is more practical than existing technologies.

One of the members of the research team noted that they “… will focus [their] future work on extending the sensor's capabilities to measure not only force along the three major axes but also rotational forces such as twist, which could be experienced during screw fastening while remaining accurate and compact. These advancements can enable the sense of touch for more dynamic and agile robots in manipulation tasks, even in human-robot interaction settings, like for patient rehabilitation or physical support of the elderly.”

Future developments of the L3 F-TOUCH sensing system could enable the development of more advanced and human-like robots one day.