TacShade 2.0: A New Greyscale Pattern-Based Optical Tactile Sensor for Robotic Perception and Adaptive Admittance Control

Marker-based optical tactile sensors generally represent contact features through the displacement of discrete markers, but their performance is constrained by marker density and potential overlap of skin wrinkles. To address this issue, we develop TacShade 2.0, a greyscale pattern-based optical tactile sensor, inspired by the sketching and halftone principles introduced in our previous work [1]. TacShade 2.0 reflects external interactions through changes in the area and density of markers. This design enables higher spatial resolution, with a density of 64 markers/cm² and utilizes continuous greyscale features to reconstruct detailed point clouds of contacted objects. A fast tactile estimation model is developed by combining local and global greyscale features to extract contact depth and normal force through pattern optimization and sensor calibration. The experiments through seven indenters reflect that the estimation errors for contact depth and normal force are as low as 0.269 mm and 0.089 N, respectively, with an average processing time of about 0.035 s per image. Furthermore, we introduce, for the first time, a tactile feedback-based variable admittance control framework. Building upon an impedance model that captures the passive compliance of the sensor, a variable stiffness coefficient is derived to minimize the force tracking error relative to the desired force. As a result, this enables adaptive robotic exploration of different surfaces. Three experiments are conducted to demonstrate TacShade 2.0’s abilities in accurate normal force estimation, fast shape reconstruction, and robust curved surface exploration, even in the presence of unknown obstacles.