Non-Matrix Tactile Sensors: How Can Be Exploited Their Local Connectivity For Predicting Grasp Stability?

Tactile sensors supply useful information during the interaction with an\nobject that can be used for assessing the stability of a grasp. Most of the\nprevious works on this topic processed tactile readings as signals by\ncalculating hand-picked features. Some of them have processed these readings as\nimages calculating characteristics on matrix-like sensors. In this work, we\nexplore how non-matrix sensors (sensors with taxels not arranged exactly in a\nmatrix) can be processed as tactile images as well. In addition, we prove that\nthey can be used for predicting grasp stability by training a Convolutional\nNeural Network (CNN) with them. We captured over 2500 real three-fingered\ngrasps on 41 everyday objects to train a CNN that exploited the local\nconnectivity inherent on the non-matrix tactile sensors, achieving 94.2%\nF1-score on predicting stability.\n