A Novel Grip Force Cognition Scheme for Robot-Assisted Minimally Invasive Surgery

Obtaining a grip force in robot-assisted minimally invasive surgery (RMIS) has always been a crucial research topic. A novel grip force cognition scheme is proposed in this article. We utilize the dynamic analysis of the cable-driven system to conduct feature engineering and add prior knowledge to the Gaussian process regression (GPR). Specifically, the feature and GPR model candidates are acquired by analyzing the dynamic characteristics first. After that, a wrapper searching method combined with exhaustive search (ES) and sequential backward selection (SBS) is proposed to determine the features and GPR model. The hyperparameters are optimized by differential evolution (DE) by minimizing the negative log marginal likelihood (NLML). Besides, the training set is extended by including several different objects to avoid the algorithm "learning" the object property. Our method is verified on a cable-driven platform to ensure the errors are all derived from the algorithm rather than the cable tension loss. Our method reduces the errors on our testing set by about 64% and outperforms several popular model-based and learning-based methods. This article illustrates the significance to integrate the system's characteristics when introducing machine learning techniques into robot systems.