eduzhai > Applied Sciences > Engineering >

TactileSGNet A Spiking Graph Neural Network for Event-based Tactile Object Recognition

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: Tactile perception is crucial for a variety of robot tasks including graspingand in-hand manipulation. New advances in flexible, event-driven, electronicskins may soon endow robots with touch perception capabilities similar tohumans. These electronic skins respond asynchronously to changes (e.g., inpressure, temperature), and can be laid out irregularly on the robot s body orend-effector. However, these unique features may render current deep learningapproaches such as convolutional feature extractors unsuitable for tactilelearning. In this paper, we propose a novel spiking graph neural network forevent-based tactile object recognition. To make use of local connectivity oftaxels, we present several methods for organizing the tactile data in a graphstructure. Based on the constructed graphs, we develop a spiking graphconvolutional network. The event-driven nature of spiking neural network makesit arguably more suitable for processing the event-based data. Experimentalresults on two tactile datasets show that the proposed method outperforms otherstate-of-the-art spiking methods, achieving high accuracies of approximately90 when classifying a variety of different household objects.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×