Haptic feedback for flexible endoscopic surgical robot using data-driven methods

• Main Author: Li, Xiaoguo
• Other Authors: Phee Soo Jay, Louis
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Mechanical engineering::Robots
• Online Access: https://hdl.handle.net/10356/152730
• Institution: Nanyang Technological University

Robot-assisted Natural Orifice Transluminal Endoscopic Surgery (NOTES) has been an emerging field of application in recent years and has demonstrated great potential and reliability in performing operations inside the peritoneal cavity while avoiding the necessity of abdominal incisions. It offers benefits such as enhanced operation precision, minimized tissue damage, and easier recovery for the patients. To actuate the joints of end-effectors through narrow and tortuous paths, tendon-sheath mechanism (TSM) is widely adopted for flexible endoscopic robotic systems. However, the friction of TSM introduces nonlinearity and backlash hysteresis which degrades the control precision and creates hurdles for developing haptic feedback. It is difficult to mount force sensors on small end-effectors due to space limitation, wiring, and sterilization issues. Previous techniques for modeling the tendon-sheath system force transmission are associated with problems such as discontinuity when the system operates at the vicinity of zero velocity and complex ad-hoc parameter identification process. This study proposes a deep learning approach to predicting the distal force of TSMs based on proximal-end measurements. A TSM-driven robotic system manipulating biological tissue was developed to collect training and testing data for deep learning. A two-stage data-driven method was developed to make dynamic distal-end force prediction of a flexible endoscopic robot without prior knowledge of its configuration. In stage one, a convolutional neural network is used to estimate the sheath cumulative bending angle based on the proximal-end force responses of the robot to a probing signal; in stage two, a combination of two long-short-term-memory models pre-trained for the bending angles nearest to the estimated angle (obtained in stage one) makes dynamic estimations of the distal-end force of the robot. The proposed approach overcomes the challenges due to unknown TSM configurations and can robustly identify the correct force hysteresis phases of TSMs. The force prediction is continuous, robust, and has a mean RMSE of 0.1711 N. This method was validated on an actual flexible surgical robot.