A Robotic Tactile Excavation System for Excavating Objects Buried in Granular Materials

Abstract

Robotic excavation of objects buried in granular materials (GM) is challenging due to the lack of visual information. In this work, we present a robotic tactile excavation system (RTES) for exploring and excavating unknown objects buried in GMs. We first design an underactuated tactile finger with multimodal tactile sensing, including sensing contact pressure and fingertip motion. A contact detection model is established to identify GMs and buried objects in real-time based on the multimodal sensor data. Then, a 3-D occupancy map of GMs is constructed to guide the object grasping. Finally, a tactile excavation (TE) policy is proposed to explore and excavate buried objects, which mainly includes two rounds of rough excavation and one round of fine excavation. Experiments on contact detection, shape detection, and TE are conducted to evaluate the effectiveness of RTES, where RTES can successfully excavate various unknown objects from GMs.