Abstract

The dataset contains the data related to three different types of data acquisitions, on which we trained and tested an artificial neural network (ANN). The procedure for the training and testing of the ANN is realized for each combination of inflated air and vertical force levels, by means of a nested cross-validation (CV). In detail, the CV is composed by two nested loops. The first data acquisition is composed by the output of the Inertial Measurement Unit (IMU) while the robotic manipulator UR5 is pressing on its surface with a metal stick end-effector on a grid on 42 different locations (namely: the 42-locations-session); the data acquired during this process from the tactile sensor are labeled based on the Cartesian position of the robot, therefore associating the signals with 42 different classes. The second data acquisition is related to the IMU data when the robot is pressing on the tactile sensor by means of a linear-like end-effector, applying the orientations of 0o, 30o, 60o, 90o, 120o and 150o (namely: the 6-orientations-session); in this case, the signals are labeled according to 6 classes, that corresponds to the six orientations of the linear region of contact points. Finally, the third data acquisition is built in the same way of the second, but considering the orientations of the linear region of contact points related to 0o, 45o, 90o and 135o (namely: the 4-orientations-session), corresponding to the labeling of the signals according to 4 classes. For each type of data acquisition, we repeated the experiment two times, and, for each of this repetition, we acquired the data for 3 levels of vertical force applied on the tactile sensor – 0.5 N, 1 N and 2 N (using the information from the force sensor at the base of the tactile sensor) – and 3 levels of inflating air – 5 ml, 7 ml and 10 ml (measured by using a syringe). In this way, we obtained a total amount of 54 datasets (27 datasets for the first session, and 27 datasets for the second session.) The data is related to the publication: Y. Iwamoto, R. Meattini, D. Chiaravalli, G. Palli, K. Shibuya and C. Melchiorri, "A Low Cost Tactile Sensor for Large Surfaces Based on Deformable Skin with Embedded IMU," 2020 IEEE Conference on Industrial Cyberphysical Systems (ICPS), Tampere, Finland, 2020, pp. 501-506, doi: 10.1109/ICPS48405.2020.9274737.