3-D Object Reconstruction Using Tactile Sensing for Multi-fingered Hand

In this paper, we propose to reconstruct 3-D model of a partially known object by tactile information while the object is grasped. Assuming the object is symmetric and its cross section can be fitted by superellipses, an initial model described in robot coordinate is generated using object's geometrical parameters and only position and orientation of object is unknown. Given the detected contacts between the fingers and the object, the initial model can be refined. This refined models will then allow the planning of more complex manipulation tasks. The main contribution of this work is an optimal estimation method for tactile data applying the symmetric object. Using little contact data the position and orientation of object in robot coordinate can be determined through Gauss-Newton iteration and Line search. The approach is validated experimentally using artificial data and real data.