Suturing Tasks Automation Based on Skills Learned From Demonstrations: A Simulation Study

TL;DR

The paper addresses automating repetitive suturing tasks in robotic surgery by developing an open-source AMBF-based simulation environment fed by an MRI-scanned phantom and a Learning from Demonstration (LfD) pipeline that combines Dynamic Movement Primitives (DMP) for both position and orientation with Locally Weighted Regression (LWR) to learn the needle trajectory. It contributes a realistic simulation setup, a standardized data collection pipeline, and a learning strategy that focuses on the needle as the learning object to improve generalization across phantom variations; results show overall generality of 0.811 and 0.915 for experienced users, with a 95% success rate in trajectory regeneration and about a 20% reduction in task time on more complex phantoms. The work demonstrates the viability of simulation-based demonstrations for suturing automation and provides a path toward deployment on physical dVRK systems, including potential skill assessment and domain transfer enhancements. Overall, the methodology advances autonomous suturing by leveraging DMP/LWR-based learning of needle trajectories and a realistic, flexible simulation environment to enable broad generalization and efficient task execution.

Abstract

In this work, we develop an open-source surgical simulation environment that includes a realistic model obtained by MRI-scanning a physical phantom, for the purpose of training and evaluating a Learning from Demonstration (LfD) algorithm for autonomous suturing. The LfD algorithm utilizes Dynamic Movement Primitives (DMP) and Locally Weighted Regression (LWR), but focuses on the needle trajectory, rather than the instruments, to obtain better generality with respect to needle grasps. We conduct a user study to collect multiple suturing demonstrations and perform a comprehensive analysis of the ability of the LfD algorithm to generalize from a demonstration at one location in one phantom to different locations in the same phantom and to a different phantom. Our results indicate good generalization, on the order of 91.5%, when learning from more experienced subjects, indicating the need to integrate skill assessment in the future.

Figures (10)

• Figure 1: Learning from Demonstration pipeline
• Figure 2: Simulation environments
• Figure 3: 3D-MED soft tissue suture pad
• Figure 4: Teleoperation setup
• Figure 5: Marker indices for phantoms