Event

SUSTech team publishes research on 10,000-point artificial skin

1 reports · 1 sources

Researchers from the Southern University of Science and Technology and collaborating institutions published research on NI-skins artificial skin in Nature Machine Intelligence: the skin reads more than 10,000 sensing elements per second and processes signals hierarchically according to a neural hierarchical architecture, taking about 23 milliseconds from sensing to interpretation. The report states that with this, a robot can pick up a 0.5-millimeter pencil lead and avoid needle pricks. The research shows that after 268-fold data compression, the system can still distinguish 10 categories of everyday objects with 98.2% accuracy. However, the current encoding-decoding circuitry is still relatively bulky, and the team proposes integrating it into a custom chip so that the electronic components that interpret touch are closer to the flexible skin.