An Effectiveness Study Across Baseline and Neural Network-based Force Estimation Methods on the da Vinci Research Kit Si System

Robot-assisted minimally invasive surgery enhances sur- gical precision and brings benefits to patients. For in- stance, the da Vinci Surgical System (Intuitive Surgical Inc., CA) enables teleoperation, which improves sur- geons’ ergonomics [1]. However, compared to a manual laparoscopic operation where the forces are transmitted directly, the da Vinci (except the newest version) lack a force-sensing mechanism. In our previous works, we have addressed this challenge with data-driven, neural network (NN) based torque prediction and force estimation meth- ods [2], [3] based on the da Vinci Research Kit (dVRK classic), an open-source mechatronic system built upon the retired first-generation da Vinci system [4]. Moreover, Yilmaz et. al. have demonstrated the potential of transfer learning, as an identification tool across different dVRK classic robots, tools, and configurations [5]. In this study, we further investigate the robustness and generalization ability of this method, using the new generation of dVRK, the da Vinci Research Kit Si (dVRK-Si). To evaluate our method’s performance, we compare the force estimation accuracy with several baseline methods. We assess and benchmark the effectiveness and transferability of these approaches by conducting comparative studies between the dVRK classic and dVRK-Si systems. This comparative study is a step towards a force estimation pipeline for various downstream applications and tasks, such as enabling haptic feedback or estimating force applied to tissue during surgery [6].