Toward Intelligent Adhesive Fibrillary Gripping Systems: Soft Coil-Based Misalignment And Shape Sensing

Bioinspired microstructured gripping systems for robots are promising solutions for future pick-and-place applications in industrial plants thanks to their energy self-sufficiency. However, the absence of a control variable limits auto-adaptive automation. To overcome this limitation, we present a coil-based soft sensor capable of detecting angular misalignment during continuous motion, as well as object shape discrimination between planar and convex shaped objects. The sensing concept is first demonstrated on activated adhesive structures, demonstrating a transition from laboratory to application. The system's angular detection performance is evaluated across a test matrix covering step-by-step and continuous robot operation, multiple velocity levels (e.g. 2mms-1, 5 mms-1, and 10 mms-1), and both TRF and WRF. Across all tested configurations, a maximum RMSE of mathematical equation is achieved. For object shape classification, a kNN classifier is trained and validated using a leave-one-group-out cross-validation, where each group corresponds to one measurement series of a single object class. Accuracy for different groups range from 97 % to 100 %. After applying majority voting with a 99 % confidence threshold 494 out of 495 measurements are correctly classified. Future work will extend the sensor evaluation by incorporating additional parameters such as contact area and pressure.