IntelliMan_WP5_Grasping, Manipulation and Arm-Hand Coordination_T5.1_Data fusion and sensing technology_Elsevier_FutureGenerationComputerSystems2023_Data

This dataset contains tactile signals acquired using two arrays of eight P(VDF-TrFE) piezoelectric sensing units integrated into the gripper of a Baxter robotic platform during controlled object-grasping experiments. The experiments involved 3D-printed objects with cubic and cylindrical geometries and different hardness levels. The "Dataset" folder contains tactile signals extracted from the grasping peaks, organized by object geometry and hardness class. Two datasets are provided for each geometry, containing 40 or 80 consecutive samples per grasping event. Each object was subjected to 170 grasping trials, resulting in 170 observations per class. The dataset includes five hardness classes (0–4) for cubic objects and three hardness classes (0–2) for cylindrical objects. Classes 0, 1, and 2 of the cylindrical objects correspond to the respective classes of cubic objects, enabling the investigation of hardness classification across different object geometries. The dataset was used to investigate machine learning approaches for real-time object hardness classification on resource-constrained embedded systems. The dataset is associated with the following peer-reviewed publication, which provides a detailed description of the tactile sensing system, experimental setup, data acquisition and preprocessing methodology, and embedded machine learning classification framework: Y. Amin, C. Gianoglio, and M. Valle, "Embedded real-time objects’ hardness classification for robotic grippers," Future Generation Computer Systems, vol. 148, pp. 211–224, 2023. DOI: https://doi.org/10.1016/j.future.2023.06.002 Users of this dataset are kindly requested to cite the above publication when using the data in their research.