Simultaneous Tactile Localization And Reconstruction of an Object During Robotic Manipulation

This paper addresses the problem of simultaneous reconstruction and localization of an object during its manipulation with a robotic hand equipped with tactile sensors. Robotic applications commonly use vision systems to perceive the environment. However, vision perception suffers from several limitations especially in manipulation tasks where objects are often occluded by the robotic hand. Tactile perception can be a good alternative to get information about the object. The tactile sensing modality is different from vision. Sensing actions may cause the object to move from its known location. Furthermore, the perceived contact points may not be exactly estimated due to the limited precision of the sensors. In this situation where the measurements and the state of the object are uncertain, a probabilistic framework is suitable for the estimation of the shape and the pose of the object. In this paper, we propose a combination of a probabilistic method for shape estimation called Gaussian process implicit surface (GPIS), and a particle filter algorithm to perform the reconstruction of an unknown object. This takes inspiration from the SLAM problem, where a mobile robot has to get a map of its environment using noisy proprioceptive and exteroceptive information.