Facebook AI Creates Realistic Robot Skin for Enhanced Touch
The Future of Robotics: Enhanced Tactile Sensing
Facebook AI Research posits that forthcoming robotic systems will demonstrate significantly improved tactile capabilities. This doesn't imply the development of robotic emotions, but rather a heightened ability to utilize the sense of touch. To facilitate progress within this emerging field of artificial intelligence and robotics, the company, alongside its collaborators, has engineered a novel electronic skin and fingertip design.
Addressing the "Why" of Robotic Skin
The rationale behind Facebook's investment in robotic skin was proactively addressed by AI head Yann LeCun during a presentation of the new projects. He recounted a query from Mark Zuckerberg regarding the justification for the company’s robotics research.
LeCun viewed this as an impetus to explore the area, and a clear purpose soon materialized. If Facebook intends to develop intelligent agents – a goal shared by most major technology companies – these agents must possess an understanding of the world that extends beyond visual and auditory input.
Beyond Vision and Sound: The Need for Physical Understanding
The sense of touch is not particularly useful for distinguishing between images of cats and dogs, or identifying speakers within a room. However, for robots and AIs designed to interact with the physical world, a more comprehensive understanding is essential.
Roberto Calandra, a research scientist at FAIR, explained, “We have become proficient in interpreting pixels and appearances, but a complete understanding necessitates a physical grasp of objects.”
The Challenge of Affordable Tactile Technology
While cameras and microphones are readily available and efficiently processed, the same cannot be said for touch sensors. Advanced pressure sensors are not commonly found in consumer products, restricting their use primarily to laboratories and industrial applications.
Introducing DIGIT: A Camera-Based Approach
Released in 2020 as an open-source design, DIGIT employs a miniature camera focused on the tactile pads to generate a detailed image of the contacted object. The sensitivity of these fingertips is evident in the detailed maps they create when interacting with various surfaces.
Image Credits: FacebookThe DIGIT project builds upon earlier work, including the 2014 MIT project called GelSight, which was revisited in 2020. GelSight is now a manufacturing partner for this well-documented tactile sensing method. (Update: Initial descriptions conflated DIGIT and GelSight with ReSkin; this section has been revised for clarity. Both DIGIT and GelSight utilize camera-based methods, while ReSkin employs a magnetic particle approach.)
ReSkin: Leveraging Magnetic Particles
The ReSkin system utilizes a different methodology. It incorporates magnetic particles suspended within a pliable gel surface. A magnetometer positioned beneath the surface detects the displacement of these particles, converting these movements into precise force maps representing the applied pressures.
Advantages of the ReSkin Design
A key benefit of the ReSkin system is the separation of its hard and soft components. The chip containing the magnetometer and processing logic is distinct from the flexible pad embedded with magnetic dots. This allows for easy replacement of the surface if it becomes damaged or soiled, while protecting the sensitive electronics.
With ReSkin, multiple chips can be interconnected in various configurations, covered with a layer of magnetic elastomer, and integrated to provide comprehensive touch information. While calibration is required, this approach simplifies the creation of artificial skin systems beyond a few square inches.
Applications and Potential
The technology can even be adapted for use in applications like pressure-sensitive dog shoes.
Animated image of a good dog with pressure-sensing pads on its feet and the readings from them. Image Credits: FacebookA pressure-sensitive surface enables robots and devices to detect objects and obstacles more effectively, reducing reliance on friction-based detection. This could lead to more gentle and responsive assistive robots – though their prevalence remains limited.
A primary reason for this limited adoption is the lack of trust in their ability to avoid causing harm, stemming from a deficient sense of touch!
Accessibility and Future Research
Facebook’s contribution lies not in groundbreaking concepts, but in enhancing the accessibility and affordability of existing approaches. The release of the software framework and the relatively low cost of the devices will encourage broader participation in this research area.
