V-ANFIS for Dealing with Visual Uncertainty for Force Estimation in Robotic Surgery

Abstract

Accurate and robust estimation of applied forces in Robotic-Assisted Minimally Invasive Surgery is a very challenging task. Many vision-based solutions attempt to estimate the force by measuring the surface deformation after contacting the surgical tool. However, visual uncertainty, due to tool occlusion, is a major concern and can highly affect the results' precision. In this paper, a novel design of an adaptive neuro-fuzzy inference strategy with a voting step (V-ANFIS) is used to accommodate with this loss of information. Experimental results show a significant accuracy improvement from 50% to 77% with respect to other proposals.