Object Pose Detection Through Touch Sensing

In this paper, we introduce a novel algorithm designed for robust assembly using only touch sensing. Visual systems are often not applicable due to various challenges, such as the object being fully or partially obscured from the camera, inadequate lighting conditions, or the camera's size preventing assembly in tight spaces. In our research, we assume the availability of a 3D map of the object, but the precise pose of the object remains unknown. Additionally, we assume the object is not flat. We developed an algorithm to efficiently address these challenges, which aims to match touch sensing results with the object's shape. Additionally, we provided extensions which efficiently handle uncertainties in the object map and touch sensing. The efficiency of the proposed algorithm is demonstrated through illustrative simulated use case and real robot experiments.