IntelliMan_WP5_Grasping, Manipulation and Arm-Hand Coordination_T5.1_Data fusion and sensing technology_IEEENEWCAS2025

Artificial tactile sensing systems have received significant attention in recent years due to their potential to improve human-machine interaction. Extracting contact forces is relevant for haptic perception and to support control actions. However, static force reconstruction is still a challenge for distributed sensing systems where sensors have a large bandwidth and do not measure static information. In this paper, we propose a machine learning based method for estimating the peak normal force in an indentation action on a tactile sensing system based on P(VDFTrFE) poly(vinylidene fluoride trifluoroethylene) piezoelectric polymers. A set of statistical time features is extracted from the raw signal acquired during the indentation process. These features are then used to train and evaluate an artificial neural network. Experimental results demonstrate the effectiveness of the proposed approach for force reconstruction, achieving a correlation coefficient R of 0.92247 and a mean absolute percentage error (MAPE) of 16.41%.