Adaptive Hybrid Control for Backlash-Like Hysteresis and Marker-Based Pose Estimation in Endoscopic Robots

Abstract

Nonlinear backlash-like hysteresis in tendon-sheath mechanism based endoscopic surgical robots introduces significant motion errors, limiting precision in surgical tasks. Existing methods struggle to compensate for these errors in real time, particularly without relying on distal-end sensors. To address this challenge, we introduce novel hybrid control strategy integrating adaptive feedforward and feedback control, enabling real-time deadband compensation and disturbance rejection. Unlike conventional approaches, our method does not require distal sensors and incorporates a vision-based tracking technique inspired by medical dye techniques. Two blue markers, attached to the robot arm, are tracked in HSV color space, allowing accurate angle estimation via polynomial regression, eliminating the need for high-clarity imaging as required in conventional laparoscopic tracking methods. Not only does our method guarantee ultimate boundedness from a control-theoretic perspective, but experimental validation under simulated surgical conditions also demonstrates a 98.6% reduction in RMSE, achieving 0.025 mm tracking error, compared to no control (1.815 mm), P-control (0.374 mm), and adaptive feedforward control (0.152 mm). The system adapts rapidly to abrupt tension changes, ensuring stable and precise control. By improving control accuracy and eliminating the need for distal sensors, our method enhances the reliability of endoscopic robotic systems, contributing to safer and more precise minimally invasive surgical procedures.