Data-driven slippage detection via tactile sensing for stable enveloping-grasp control under external disturbances

Enveloping-grasp control using tactile sensors for slippage detection has garnered significant attention in recent years. However, it remains an open problem in practical applications. This work proposes a data-driven slippage identification framework to address the challenge of enveloping-grasp control. The framework employs a tactile sensor array composed of micro-electro-mechanical barometers covered with soft material, which perceives contact behavior as spatiotemporal data. In the absence of detailed knowledge of interaction dynamics, the slippage information is estimated using a tactile learning model with convolutional computations. The slippage identification output is integrated into an enveloping-grasp controller to resist slippage. Compared to previous studies, our method is cost-effective and does not require large amounts of data or high computational power. Extensive experiments validate the effectiveness and generalization capability of the proposed framework, demonstrating that the slippage identification method exhibits high classification accuracy for grasp status of both known and unknown objects.