Xiaowei Shi

Shape recognition is essential in multifarious manipulations, such as object grasping, block building, and parts assembly, which most frequently relies on visual perception. However, it inevitably suffers from adverse impact of illumination intensity, visual distortion from different views, and frequent occlusions in dynamic scenes. To tackle this problem, the paper proposes recognizing different shapes via tactile exploration, mimicking the human manner of touch in dark environments. Based on our designed tactile fingertip, which comprises four sensing elements and a fingerprint structure, we employed the signal processing framework consisting of data preprocessing, feature pool establishment, selection of the most discriminative feature combination, and recognition model using a shallow neural network. Particularly, the feature pool was extended to comprise 386 multi-perspective features spanning different domains and scales. When the nine features with the highest weights were selected using Neighborhood Component Analysis, our extreme learning machine-based recognition model achieved an accuracy as high as 99.72%. To contextualize this performance, a comprehensive benchmark was conducted against several state-of-the-art deep learning paradigms, including end-to-end and transfer learning models. The analysis reveals that for data-constrained and resource-sensitive tactile sensing applications, our proposed and interpretable feature engineering approach not only achieves superior accuracy but also offers significant advantages in computational efficiency. This work validates the practical value of tailored feature engineering and sheds light on more efficient utilization of tactile signals in real-world robotic recognition tasks.

Pressure sensors, especially the typical capacitive sensors that feature low power consumption, have drawn considerable interest in emerging and rapidly growing fields such as flexible electronics and humanoid robots, but often suffer from limited performance. Here, we report a contact-dominated design for capacitive pressure sensors to dramatically improve the sensing response and linearity over a broad pressure range. This design is implemented by utilizing hierarchical microstructured electrodes made of robust conductive composites with metallic coverage and layered dielectrics with high unit-area capacitance to realize localized electric-displacement-field-enhanced capacitance change. We demonstrate a significant improvement in pressure response beyond 3000 and a sensing range exceeding 1 MPa, particularly with a near-linear response (optimized R² of 0.9998) and high sensitivity of 9.22 kPa⁻¹ in a wide pressure range of 0–100 kPa. Moreover, we present that the integration of the contact-dominated sensor with floating-gate low-dimensional semiconductor transistors can provide a transduced electrical response of ~4 × 10⁵ at a low operating voltage of 2.66 V due to the greatly enhanced pressure response. We also demonstrate the potential applications of our sensor in fluid physical property evaluation and precise dynamic control of a robotic arm for manipulation tasks.

To date, numerous tactile sensors and algorithms have been developed to tackle various perception problems. However, excessive endeavors are devoted to the multifarious definition, extraction, and analysis of hand-crafted features in order to improve the perception accuracy. To address this problem, in this paper, we designed a tactile finger containing four sensing elements to perceive both dynamic and static stimuli, and meanwhile proposed a novel signal processing pipeline. The pipeline mainly consisted of time-series signals conversion, automatic deep features extractor, and a shallow recognition model. When the tactile finger was applied to explore 16 surfaces based on a robotic platform, the four-channel signals were converted and concatenated into a time-frequency image via continuous wavelet transform. A deep feature extraction network was constructed based on a pretrained deep learning (DL) model, Resnet101, to extract the required features, which acted as high-level representations of the most discriminative components from the tactile images. Finally, these features were fed into a shallow machine learning model, i.e., extreme learning machine, achieving an accuracy as high as 92.38%. In such a manner, the powerful learning capability of DL models was transferred to the new recognition model directly while the tedious feature extraction procedures were alleviated. Besides, several relevant issues, such as the layer depth, the DL model type, and the shallow recognition model, are addressed and discussed to reveal their influences on the performance.

With the endowed ability to perceive various objects on touch, human is capable of manipulating them dexterously and interacting with surrounding environments, which inspires a huge number of tactile sensors to spring up. In this paper, we designed an improved tactile fingertip for robotic manipulation. Bio-inspired by the fast-adapting (FA) and slow-adapting (SA) types of mechanoreceptors within human skin, the fingertip contains two polyvinylidene fluoride (PVDF) films and two strain gauges as sensing elements, enabling the perception of both dynamic and static forces. Meanwhile, the structure of human fingertip is mimicked in the design that sensing elements are covered by soft ’skin’ while a stiff bar plays the role of phalanx. To realize one of the typical haptic abilities, surface recognition, an efficient signal processing framework is proposed to focus on the extraction and selection of the most discriminative features rather than paying excess attentions to various feature definitions. The selection algorithm, neighborhood component analysis, is combined with a linear discriminant classification model. Through a robotic experiment, effectiveness of the tactile fingertip as well as the recognition method is validated, aimed at recognizing ten object surfaces in our daily life. The final recognition accuracy reaches as high as 99.52%. Furthermore, several typical algorithms for feature selection and object classification are compared while the influence of dataset parameters on the prediction performance is investigated.