An Optical Artificial Skin for Pressure Point Localisation on Robotic Arm Links

This paper presents the development and evaluation of an optical artificial skin capable of accurately measuring contact location and pressure. The artificial skin consists of a flexible transparent gel that is wrapped around on a cylinder such that it can cover robotic arms and prostheses. The setup incorporates pairs of phototransistors and light emitting diodes placed at opposite ends of the cylindrical structure. The investigation uses white and infrared light sources and two types of contact point shapes. The data capturing is automated to improve efficiency and reduce errors. Neural Networks are employed to capture the physics of light propagation in the gel material. Using white light, the system achieves a Root Mean Squared Error (RMSE) of 2.7 cm in determining the contact point position, correctly identifies the contact point shape 95.2% of the time and correctly classifies the pressure level 88.5% of the time. With infrared light, the RMSE for the contact point position is 3.65 cm, the contact point shape is identified correctly 96% of the time, and pressure level classification is correct 85.6% of the time.