Task characterization for intelligent control of tendon-driven serpentine surgical manipulators with learning from demonstration

Dr Lau, Henry Ying Kei   (Principal Investigator (PI))

30

2018-01-01

295966

Task characterization for intelligent control of tendon-driven serpentine surgical manipulators with learning from demonstration

automatic control, motion planning, robot task characterization, serpentine robot, skill learning

Robotics and Automation,Design

Engineering (E)

17204717

General Research Fund (GRF)

2017

Completed

1) To develop a motion characterizing framework to generate motion primitives for the control of a tendon-driven mechanism based on typical motion commands from a surgeon performing MIS; 2) To develop a kinematics free and sensorless learning from demonstration (LfD) method to automate the performance of surgical operations via actuation space learning towards automation of surgical tasks for robot-assisted MIS. The accuracy of the method is enhanced by data-driven methods; 3) To build a dual arm prototype surgical robot each consisting of two segments. One of the arms is equipped with an endoscope, and the other one is installed with surgical instrumentation; 4)To form motion primitives and the human robot interaction (HRI) library by identifying benchmark surgical movements from surgeons, including cutting, targeting, and segment such movements into motion primitives, then LfD is used to encode these primitives with parametric models. The learned models will be stored in the HRI library which can be used and adjusted by surgeons to automate part of the surgical interventions and facilitate supervised tele-surgery; 5) To obtain performance data of the proposed control strategy through experimental studies with the prototype surgical robot to perform a number of representative surgical interventions, including suturing and penetration, and compare the performance with typical manual operations.