Year: 2023
Type: dissertation
Abstract: Continuum robots are a type of hyper-redundant manipulator, which is generally structured with multiple small diameter sections and can access and operate in narrow & hazardous environments. This research project aims to explore the shape-sensing technology and feasible control strategy for a multiple-section continuum robot system, to solve the control challenge introduced by the hyper-redundancy (i.e., the number of actuators is less than the number of DOFs) and coupled motion (i.e., for the robot with more than one section). Through this work, a hybrid sensing system is developed and implemented in a prototype continuum robot, with a two-level closed-loop controller developed and a vision-aided led-by-the-tip navigation strategy proposed based on it. This sensing solution consists of a dual displacement sensor system and a self-carried visual tracking system. The former works to measure the cable displacement error at the actuation pack and predict the cable elongation in the arm structure. The latter equips the robot with several camera-marker arrays along the arm and a head camera at the distal tip, providing section-by-section tip tracking and visual perception for the robot controller. Integrated with the two-level closed-loop controller, the whole system is designed to minimize control error in both joint-space and task space. Further, an adaptive led-by-the-tip navigation strategy is proposed to navigate the robot arm into confined space, with a head camera observing the working environments. Under this work frame, the robot can snake its way through narrow channels and perform environmental interactive tasks such as obstacle avoidance and object detection. All the works were demonstrated on a two-section cable-driven continuum robot with a rotary actuation system design, mounted in a linear rail to provide feed-in motion. Each section is 120 mm long and allows 2-Dof motion in 3-D space. It was shown that the measurement error of the proposed self-carried visual tracking system is smaller than 2 mm, with a mean error of 0.67 mm within the workspace of a 120 mm single section. The distal-tip control accuracy of the 2-section prototype robot can achieve 1.72 mm in the world coordinate (in the X-Y plane), which reduces the tip control error by 73.5% compared with the conventional PCC (piecewise constant curvature) model-based controller, for both smooth and sharp movements. In open-space navigation tasks, along a 240 mm guiding path, the robot achieved an average trajectory error of 3.5 mm with a tip-following deviation of 2.5 mm. Then, the robot is navigated into a mock-up with boundaries and obstacles to demonstrate its capability for online obstacle detection and motion planning. These experiments validate the feasibility and reliability of the sensing system and controller architecture proposed in this study.
Author: Fang, Yihua
Institution: University of Nottingham
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7217425283
