Learning surgical skills under the RCM constraint from demonstrations in robot-assisted minimally invasive surgery

Learning surgical skills from trained surgeons can increase the level of autonomy of surgical robots and provide assistance for surgeons in an appropriate way during surgery. However, the remote center of motion (RCM) constraint is a tricky problem while most other works only consider the task performed in the lesion area. This study aims to transfer the minimally invasive surgical skills demonstrated by surgeons to the surgical robot while satisfying the RCM constraint. In this paper, the implicit constraints of manipulation skills are modeled into a probabilistic model to maintain the variability and flexibility of the surgeon's operations. A novel method is proposed to address the inconsistency between the RCM constraint space and surgical task space. The generalization of the learned skills under the RCM constraint has also been improved. We validated the proposed method in a physical experiment with a tracking task under the RCM constraint. An original measurement method based on shape similarity is proposed to compute the tracking errors of trajectories that have nonhomogeneous temporal and spatial distortions. The root means square error of the trajectory was 1.8 mm, which exceeded the average for operator demonstrations.