are at the same position, which ensures that the whole process has no influence on subsequent tasks. A training data set consisting of 4 × 40 groups of raw sensor data for texture discrimination was collected using the robotic rat in advance. The following experiments were performed after the SVM algorithm had been applied to the training data set. We controlled the robotic rat to sweep the whisker sensor up an object vertically while the distance between the wall and base of the sensor was fixed at 5 mm. In this way, we conducted 20 sets of experiments on each of the four textures and transmitted the experimental data to the previously trained SVM classifiers. The experimental result shows that most test data groups (4 × 20) were successfully classified into the corresponding textures categories. The accuracy of the classifier on the test data set is 88.3%, and the result is detailed in Figure 6. Writing Paper Tissue Paper Flannelette Sandpaper Figure 6. The results of the texture discrimination, where a dot denotes that a prediction is correct and an x indicates that a prediction is wrong. Performance of Contour Reconstruction For the contour reconstruction experiment, two object samples with the representative shape of a semicircle and a rectangle were prepared. As can be seen in Figure 7(a) and (c), when the whisker is aligned with the y-axis, only 1 DoF in the x-axis driven by the robotic rat is variable. An object sample with a semicircular shape and diameter of 25 mm was used for the first contour reconstruction experiment; it represents objects with contour curvatures greater than zero. The robotic rat Enlarged View 35 30 25 20 15 10 05 10 (a) Enlarged View 35 30 25 20 15 10 05 10 (b) Figure 7. (a) and (c) The contour scanning (at 20 mm/s) for objects with a semicircle contour and rectangular contour, respectively. The parts of the whisker are enlarged, and the whiskers are drawn thicker for emphasis. (b) and (d) The reconstruction performance. 122 * IEEE ROBOTICS & AUTOMATION MAGAZINE * DECEMBER 2022 x (mm) Object Contour Reconstructed Shape Error Based