An S-SOM method based on binocular vision for configuration detection of concentric cable-driven manipulators

Abstract The concentric cable-driven manipulator (CCDM) is slender enough to work in narrow oral environments. However, the configuration detection of the CCDM with two segments is the key issue to realize the servo control and force sensing. In this paper, a Segmented Self-Organizing Map (S-SOM) method is presented for configuration detection of the CCDM without preset markers. Firstly, two-dimensional point cloud of the CCDM is acquired by the binocular vision. Secondly, the dimension reduction of point cloud is realized by principal component analysis to avoid noises and eliminate redundant information. Then, the threshold is set to distinguish the segment with different diameters that need to be trained independently. Thirdly, quantization errors and topological errors are designed to select different training parameters of the CCDM with two segments. Finally, the experimental results show that the average point-to-point distance error is less than 2.57 mm. The effectiveness of the proposed S-SOM method for configuration detection of the CCDM is verified.