Statistical modeling on motion trajectories for robotic laparoscopic surgery

Abstract

Learning by demonstration enables a robot to learn and perform tasks from kinesthetic demonstrations. Gaussian mixture method with constraints is applied in this work to model the motion using its trajectories and enable a robot to learn motion skills for a simple surgical task with specific requirement. Tissue dividing experiments are demonstrated on a robotic surgical simulation platform to collect motion trajectories. The demonstrations are modelled using Gaussian Mixture Model. Constraints are also imposed onto the motion model to suit the specific requirements for carrying out the surgical task on a virtual patient. The robot is demonstrated to be able to learn the surgical skills with the statistical model and execute it to complete a virtual surgical task.