A Real-Time, Semi-Autonomous Navigation Platform for Soft Robotic Bronchoscopy

Navigating through the peripheral lung branches poses a significant challenge in diagnosing lesions during bronchoscopy. Soft robots are well-suited to address current limitations in bronchoscopy due to their scale, dexterity, and adaptability. In this paper, we propose a real-time, semi-autonomous navigation platform that leverages a soft continuum robot with an outer diameter of 2.5 mm for tip steering and a UR5e robot arm for insertion, translation, and rotation. Closed-loop feedback is provided via on-board visualization and electromagnetic tracking. Steering capability and workspace are characterized to demonstrate sufficient robot tip dexterity. A driving algorithm combined with a YOLO-based computer vision algorithm is developed to enable the robot to steer toward the target branch along preplanned paths. Multiple successful navigational experiments were performed within an in-vitro lung phantom to validate the proposed platform. The scale of the robot allows for successful navigation deep into the smaller, peripheral branches of the lung (6th generation) and exits the lung phantom, demonstrating the ability to reach the lung periphery with an average error at the target location of 1.1 mm.