Shape reconstruction of soft continuum robots via the fusion of local strains and global poses

Soft continuum robots are gaining increasing attention for their high dexterity, and shape reconstruction is one of the key techniques. Here, we propose a shape-reconstruction method through the fusion of locally measured strains and poses to eliminate spatial error accumulation when estimating the shape relying solely on strains. We transform the shape reconstruction into an optimization problem with the goal of minimizing the difference between the measured and predicted orientation, where the constraints are set as local bending curvatures. We implement our method on a custom-made robot with two types of embedded sensors-a multi-core fiber Bragg grating sensor and three inertial measurement units. Results show that, compared with the conventional piecewise constant curvature method, our method exhibits a remarkably higher shape-estimation accuracy and robustness to strain errors. We believe our method and the embeddable nature of sensors pave new ways for robots' shape sensing in practical applications.