Modelling, analysis and control of tendon-driven continuum robots

Continuum robots (CRs) outperform the conventional rigid-link industrial manipulators owing to their hyper-redundant and compliant properties, enabling them the capability of accessing to confined environments (e.g., aero-engines and pipeline systems) and applications requiring flexible interactions such as minimal invasive surgeries. As such, an Innovate UK project was setup to validate the concept of utilizing slender continuum robots for the in-situ maintenance of aero-engines. This research firstly follows the development of the control algorithms for the continuum robot constructed in the project, which is a 715 mm long continuum robot with ten 55 mm long one-DoF (Degree of freedom) body-sections and three 55 mm long two-DoF sections at the tip, together with a linear stage for feed-in motion. The robot has been demonstrated within a range of aerospace scenarios. Besides, precise control of single-section continuum robots is also within the scope of this research, aiming to develop universal high-performance controllers for various designs. Through this work, model-less controllers and model-based hybrid controllers are designed for slender multi-section and single-section continuum robots, respectively. Inspired by the tug of war, a novel local model-less controller utilizing fuzzy logic algorithm is proposed for the tip sections control on the slender long continuum robot. This implements the control policies directly from the task space to the actuation space, avoiding the model mismatch of the PCC (Piecewise-Constant-Curvature) assumption owing to the explicit call of arc parameters. Experiments on a single section of the tendon-driven continuum robot, in comparison with PCC-based method, validate the stability of the developed controller, which can reach ±1 mm overall positioning accuracy and ±0.5 mm positional accuracy for 75% of the test points in both X and Y directions. Further, a set of trails on two distal sections of a long robot demonstrate that the controller can also effectively minimise the section coupling issue. In terms of single-section continuum robots, hybrid control scheme is proposed after the analysis of the regular control modes of continuum robots and the advantage of such type of robots, which enables tension supervision to avoid slack state owing to the cumulative control errors during operation. Also, a novel differential kinematics is derived directly from the task space to actuation space by using polar coordinate system to replace the Cartesian coordinate system in conventional kinematics and applying accurate piecewise linear approximation. The proposed controller is experimentally compared with the PCC-based method on the continuum robot with twin-pivot joints. The experimental results demonstrate that the proposed controller had more comprehensive stability for positioning, while both controllers show their advantages on different performance indices such as RMSE(Root-Mean-Square-Error) and maximum error in terms of path tracking. Further performance characterisation of the proposed controller is conducted on the dual-revolute and soft-robot designs