Contents I. Introduction
In robotic manipulation, grasping control is still considered challenging because of the absence of sufficient feedback, combined with uncertainties from unknown model parameters and nonlinearities caused by deforming soft objects. If robots are to emulate human hands, grasping must be adjusted while manipulating random objects, all without human intervention. For these reasons, it is necessary to modulate the degree of grasping based on high quality feedback data and using models optimized for different situations. Since human hand contact facilitates touching, grasping, and picking based on tactile sensory information, many previous researches have tried to imitate humans' soft skin features to make robots sense contact force information as well. Soft, flexible features enable hands to grasp delicate objects and conform according to surface roughness. This is the main reason why most of the recent works choose soft and flexible materials as tactile sensors. [1]–[10] Optical sensing devices made from elastomers and highly compliant polymer fibers also have been highlighted recently since they are soft, flexible, and experience less electromagnetic interference, while their performance is on par with electronic capacitor or resistor-based e-skins. [11], [12] When it comes to slipping, since the normal force is not sufficient to detect, these soft tactile sensors provide not only normal force but also lateral force information as tactile data. [13], [14] Especially, in our previous work [15], a soft-type optical tactile sensor has been introduced to obtain not only grasping normal force but also lateral friction information.
