Tactile Imprint Simulation of GelStereo Visuotactile Sensors

Visuotactile simulation is crucial for robot dexterous manipulation skill learning, but remains a challenging problem. Most of the existing methods are designed for GelSight-type sensors, which focus on optical response simulation and do not pay much attention to the contact dynamics of elastomer. In this paper, we propose a high-fidelity tactile imprint simulation model of the visuotactile sensors, which uses finite element analysis (FEM) to simulate realistic contact dynamics. Moreover, a vertex mapping-based rendering method is proposed to realize real-time textured tactile imprint rendering using finite element results, which can be further used to extract tactile features, such as markers motion field. In order to achieve high-fidelity simulation effects, we also design a simulation parameter calibration method, which calibrates the finite element parameters of the elastomer simulation by optimizing the mean square error of the distance from the corresponding markers of the contact area (Marker-MSE) between the real and simulated elastomer deformation. Furthermore, a large number of quantitative and qualitative simulated contact experiments are performed, and the results strong verity the effectiveness and robustness of the proposed simulation model. This study is an active exploration of the simulation of GelStereo-type sensors, and it has positive significance for the simulation modeling of visuotactile sensors that focus on contact dynamics and textured imprints.