A transformative tool for minimally invasive procedures: Design, modeling and real-time control of a polycrystalline shape memory alloy actuated robotic catheter

Cardiac catheterization is rapidly transforming the diagnosis and treatment of cardiovascular disease. However, the use of catheters is limited to procedures where the target anatomy can be easily accessed via natural vasculature. Robotically controlled catheters have the potential to provide greater access and more precise interaction with internal anatomies. This dissertation presents the development of a shape memory alloy (SMA) actuated robotic catheter: from electromechanical design to the development of novel modeling and control approaches.;The robotic catheter is fabricated using conventional manufacturing and rapid prototyping. To analyze the transient characteristics of the catheter, a dynamic model is developed. Its bending mechanics are derived using a circular arc model and are experimentally validated. The effects of outer sleeve thickness on heat transfer and transient response characteristics are studied. SMA actuation is described using the Seelecke-Muller-Achenbach model for single-crystal SMA with experimentally determined parameters. Joule heating is used to generate tip deflections, which are measured in real-time using a dual-camera imaging system. The dynamic characteristics of this active catheter system are simulated and validated experimentally.;The direct extension of the Seelecke-Muller-Achenbach model to a catheter with multiple SMA tendons proves difficult because of the computational cost and inherent inaccuracies of single-crystal modeling assumptions. Moreover, the requisite variable-step solvers are not suitable to real-time control. To facilitate more accurate modeling and effective real-time control of an SMA catheter with multiple tendons, a new modeling technique based on Hysteretic Recurrent Neural Networks (HRNNs) is proposed. Its efficacy is demonstrated experimentally for two- and three-phase hysteretic systems. The HRNN is shown to accurately capture the polycrystalline stress-strain characteristics of SMA tendons at different temperatures.;A robotic catheter system consisting of four SMA tendons is then decoupled into two planar bending systems, each containing a pair of antagonistic SMA tendons. An HRNN model is developed directly from experimental output measurements, and is used to develop a feed-forward controller.