Tactile sensor and algorithm to detect slip in robot grasping processes

In this paper we introduce a tactile slip sensor for an anthropomorphic robot hand, the measurement circuit and the corresponding algorithm to determine slip states. The main slip sensor components consist of a silicone rubber surface which covers a PVDF sensor. After the amplification of the signal it is processed by a discrete short-time Fourier transform. The resulting spectrogram is processed by a principal component analysis to determine the main signal components. For a classification of three states (dasiaslippsila, dasiasignal but no slippsila and dasianoisepsila) a k-nearest neighbour classifier has been trained with the main signal components and serves for discrimination of slip states on the sensor´s surface. The build-up of the sensor and the experimental setup will be briefly explained, the signal processing and the results will be discussed in detail.