EuroWire November 2020

Transatlantic cable

It is hoped that further investigation of rhombohedral graphite will elucidate on the origin of many-body phenomena in strongly correlated materials such as high temperature superconductors.

“A lot of preliminary work in other fields indicated that sound could be useful, but it wasn’t clear how useful it would be in robotics,” said Lerrel Pinto who, with colleagues, found that robots using sound had a 76 per cent success rate in classifying new objects. Professor Pinto added that the results were so encouraging “that it might prove useful to equip future robots with instrumented canes, enabling them to tap objects they want to identify.” The researchers presented their findings during the virtual Robotics Science and Systems conference. Other team members included Abhinav Gupta, associate professor of robotics, and Dhiraj Gandhi, a former master’s student who is now a research scientist at Facebook Artificial Intelligence Research’s Pittsburgh facility. To perform the study, the research team created a large dataset by simultaneously recording video and audio of 60 common objects as they were moved around a tray. The interactions were captured while using an experimental apparatus named Tilt-Bot – a square tray attached to the arm of a Sawyer robot that spent hours moving the tray in random directions with varying levels of tilt while cameras and microphones recorded the actions. They also collected data away from the tray, using the robot to push objects across other surfaces. The team found that a robot could utilise what it learned about the sound of one set of objects to make predictions about the physical properties of previously unseen objects. “I think what was really exciting,” said Prof Pinto, “was that when it failed, it would fail on things you expect it to fail on.” For example, the robot was unable to make a prediction about colours by using sound, but could differentiate between types of object such as a hollow cup and a solid building block. CMU’s dataset of 15,000 interactions is available to other researchers. AI system that can be led by the hand Scientists from Australia and Singapore have developed an artificial intelligence system that, by combining skin-like electronics with computer vision, recognises hand gestures. The team from Singapore’s Nanyang Technological University (NTU) and the University of Technology Sydney (UTS) published its findings in Nature Electronics. According to NTU Singapore, gesture recognition precision is currently hampered by the low quality of data transmitted by wearable sensors, often due to their bulkiness and poor contact with the user, or to obstructions or poor lighting. Further challenges arise from the integration of visual and sensory data as they represent mismatched datasets that demand separate processing before merging, which is inefficient and leads to slower response times. The NTU team’s bio-inspired data fusion system uses skin-like stretchable strain sensors made from single-walled carbon nanotubes, and an AI approach that mirrors the way in which the combination of skin sensations and visual information are handled by the brain. The AI combines three neural network approaches in one system: a “convolutional neural network”, which is a machine learning method for early visual processing; a multilayer neural network

Nearing room temperature superconductivity?

A broad team of Penn State and Florida State physicists and materials scientists, funded by the US Department of Energy, has made a discovery that could bring room temperature superconductivity closer to attainment. Their discovery involved layering a two-dimensional material calledmolybdenumsulphidewithanothermaterial,molybdenum carbide. A known superconductor, molybdenum carbide allows electrons to flow through it without resistance. “Superconductivity occurs at very low temperatures, close to absolute zero or 0 Kelvin,” explained Mauricio Terrones, one of the authors of the paper – “Superconductivity enhancement in phase-engineered molybdenum carbide/sulfide vertical heterostructures” – published in the Proceedings of the National Academy of Sciences.“The alpha phase of [molybdenum carbide] is superconducting at 4 Kelvin.” When layering metastable phases of molybdenum carbide with molybdenum sulphide, superconductivity occurs at 6 Kelvin, a 50 per cent increase. Though not unique – other materials have been shown to be superconductive at temperatures up to 150 Kelvin – it was still an unexpected phenomenon that suggests a new method to increase superconductivity at higher temperatures in other superconducting materials. “Calculations using quantum mechanics, as implemented within density functional theory, assisted in the interpretation of experimental measurements to determine the structure of the buried molybdenum carbide/molybdenum sulphide interfaces,” said Susan Sinnott, professor of materials science and engineering, and head of the department. “This work is a nice example of the way in which materials synthesis, characterisation and modelling can come together to advance the discovery of new material systems with unique properties.” Prof Terrones added, “It’s a fundamental discovery, but not one anyone believed would work. We are observing a phenomenon that, to the best of our knowledge, has never been observed before.” The teamwill continue its experimentation with superconductive materials with the ultimate aim of finding materials combinations that will carry energy through the grid with zero resistance. Robot perception enhanced with hearing A team from Carnegie Mellon University (CMU) believes robot perception can be improved by adding hearing to the machine’s sensing skills. Researchers at CMU’s Robotics Institute have found that sound can help a robot to differentiate between objects, and could not only help robots determine what type of action caused the sound but also allow them to use sound to assess the physical properties of a new object. Robotics and AI

38

www.read-eurowire.com

November 2020