High-Performance 3-D Liquid-Metal Tactile Sensing System With Data-Driven Signal Decoding for Dynamic Force Estimation

Tactile sensing plays a crucial role in human-robot interaction, surgical assistance, and intelligent operation. However, achieving stable and accurate perception of multidimensional contact forces under dynamic contact conditions still presents challenges in structural design and signal decoding. To address this issue, this paper proposes a three-dimensional tactile sensing system based on gallium-indium (GaIn) liquid metal for acquiring and estimating triaxial force information during dynamic contact. This sensor employs a multi-layered flexible structure and a distributed sensing unit design, enabling simultaneous sensing of normal and tangential force components. Considering the strong coupling characteristics of liquid metal tactile signals in both spatial and temporal dimensions, this paper introduces a data-driven decoding method combining convolutional neural networks and long short-term memory networks (CNN–LSTM) to reconstruct continuous triaxial contact forces from multi-channel tactile signals, achieving prediction accuracy superior to various baseline regression models. Experimental results demonstrate the feasibility of this liquid metal tactile sensing scheme in dynamic three-dimensional force estimation tasks, providing an effective tactile sensing means for precise robot manipulation and safe human-robot collaboration.