Active tactile exploration with uncertainty and travel cost for fast shape estimation of unknown objects

In this paper, active tactile exploration for object shape estimation is explored. A prior work suggested to touch the most uncertain part of the estimated shape for minimizing the required number of touches. In this paper, it is pointed out that it may not be the best approach for fast estimation. We propose a novel criterion in active touch point selection for fast estimation, which considers both uncertainty of shape estimation and travel cost to touch. Our method employs a Gaussian process implicit surface model to learn the object shape from tactile information, which allows us to evaluate the uncertainty of the shape estimation with an analytic form. To estimate the travel costs for all the touch candidates, our method utilizes a computationally-efficient graph-based path planning method based on stochastic optimal control theory. Simulations with 2D and 3D objects and real-robot experiments with a 7DOF robot arm and a single finger device equipped with a tactile sensor are conducted. Experimental results verify the effectiveness of our method for fast tactile object shape estimation as compared to passive exploration and active exploration that touches the most uncertain part of the estimated shape.