Wheeled Visuotactile Sensing for Aerial Physical Interaction With Surfaces of Unknown Curvature

Abstract

Tactile perception holds the potential to enhance the capability of aerial robots to physically interact with unknown environments. However, current tactile sensor designs primarily focus on static scenarios, limiting their applicability to dynamic interaction tasks. This article presents a novel wheeled visuotactile sensor that enables the perception of surface information during rolling interactions. A novel color pattern, along with the corresponding marker detection and matching algorithms, is developed for contact position measurement during rolling interaction via deformation reconstruction of the contact body. Based on this perceptual capability, a real-time trajectory generator integrated with an impedance/force control framework is proposed. The performance of the visuotactile sensor is assessed via tests on various color pattern designs, rotation velocities, and surface curvatures. Additionally, an aerial manipulator prototype, integrated with a two-degrees-of-freedom manipulator and the visuotactile sensor, is established to demonstrate the stable rolling interaction on a cylinder with unknown geometrical parameters.