Performance Prediction of Bio-inspired Compliant Grippers Using Machine Learning Algorithms

The utilization of compliant grippers is rapidly growing in numerous fields due to the need for greater control and safety in grasping applications. Compliance grippers’ design and evaluation process often encounter challenges such as prolonged duration and low success rate when traditional methods are employed. Additionally, the current process of designing and testing these grippers is time-consuming and often relies on a trial-and-error method or the application of finite element method (FEM) analysis. This study endeavors to enhance compliant grippers’ design and testing stage by employing machine learning algorithms (MLA). Support Vector Regression (SVR), K-Nearest Neighbor (KNN) Regression, Decision Tree Regression, XGBoost Regression, and Gaussian Process Regression are used to generate rapid and precise predictions. Two datasets are utilized, with the first one comprising experimental measurements aimed at validating the results obtained by FEM analysis, as they are subsequently used to train MLAs. The primary objective was to assess the accuracy of these MLAs in predicting gripping linear displacement based on various design parameters. Ultimately, the research revealed the MLA that exhibited the highest performance, characterized by the lowest prediction error with selecting the optimal set of design parameters.