| The adoption of robotic technology has instituted further development towards a new paradigm-shift in interventional surgery.Unlike conventional laparoscopic surgical methods,that require multi-ports for single intervention,novel robotic instruments are currently being designated for intraluminal surgery.This advancement in robotic surgery has potentials to further reduce trauma on the sides of surgeons and patients,and at the same,minimize hospitalization overheads.Recently,flexible surgical robots with snake-like,catheter-based,and continuum systems,are being developed for flexible access surgery.These advanced mechatronic designs are proposed to access operative targets through natural cavities.With these,surgical interventions can be carried out on vital anatomical organs within human body by spatial navigation of surgical effectors through flexible intraluminal pathways.Thus,surgical devices can move along anatomical pathways without necessity for wider or multiple incisions,as in conventional methods.However,the flexible robotic systems suffer underactuation-related problems.Thus,effective models are required for navigation and teleoperation through the complex intraluminal pathways.In this thesis,statistical learning-based models were developed for motion control and teleoperation of snake-like and catheter-based robots.The snake-like robot was designed for surgical treatments of intraluminal cancer in the gastrointestinal tract.The modular design has pairs of serial-links connected with orthogonal micromotor-actuated joints.Similarly,the catheter-based system includes iterative prototypes of two degrees of freedom(DoF)robotic platforms that were proposed for catheterization of flexible surgical tools in intraluminal vascular pathways.Current designs of the robotic systems have motivated developments of control models as parts of developmental projects for intraluminal robot-assisted surgery.However,underactuation is a common problem deterring precise navigation of the flexible robotic systems along spatial anatomical pathways.Thus,statistical learning-based models are proposed for motion control and teleoperation of the flexible robots for intraluminal surgery.These include technical contributions of non-iterative and iterative inverse kinematics(IK)models proposed for precise and fast kinematics resolution of the snake-like robot.The former is based on geometric analysis for estimation of virtual points that each joint in a given snake-like model can be positioned so as to have the robot's end-effector adeptly placed at targeted points.The IK models were verified with PRn~2R and 2~nR mechanisms of the snake-like robot,and better performances were achieved in terms of IK accuracy and execution time.Transcendence in geometry-based IK modeling increases with every increase in DoFs of the snake-like robot.As a result,solving IK of the robotic model with higher dexterity is very complex and solution is not always guaranteed;thus,the motivation for application of iterative IK model,in this thesis.Iterative modeling involves application of Jacobian damped least squares(DLS)method with solutions based on single and unique damping factor(s)to solve IKs of snake-like robots.This method is proposed as a deeply-learnt DLS model built to ensure consistent IK resolution of the snake-like robot.The learning was achieved with a deep neural network trained to predict adept damping factor needed to solve IK of data points in the robot's workspace.Implementation with 8-DoF prototype of the snake-like robot shows the proposed IK model has good performances in terms of higher reachability rates,minimized kinematics error,and execution times,simultaneously.Success of the kinematics methods was utilized for dynamic constraints analysis with focus on modeling both the position and generalized forces of the underactuated flexible surgical robots.The recursive Newton-Euler formulation was adapted to model the motion profiles and generalized forces at each part of the snake-like robot to ensure safe and steady motion during trajectory tracking.Simulation with 19 sec point-to-point motion dataset shows that the average execution time of the model is 21msec.Thus,the kinematics and dynamics model can support fast and very precise control of the snake-like robot during teleoperated surgery.Another major contribution of this thesis is statistical characterization and adaptive model for backlash compensation in a 2-DoF robotic catheter system.It is vital to state that the catheter-based robot is based on axial driving of flexible and underactuated surgical tools,which are stateless.Thus,IK models are inapplicable to their motion controls.However,catheterization of the tools requires analysis of non-linear hysteretic factors that constrains their motion during intraluminal vascular surgery.Thus,single factor descriptive characterization and adaptive compensation model were proposed for dynamics analysis and compensation of backlash during robotic catheterization.The adaptive system consists of a neuro-fuzzy module that predicts backlash gap based on bounded motion signals and modulated contact force from distal tip of a customized catheter.In-vitro robotic catheterization procedures were carried out with phantom model of human vascular pathway.Lastly,suitability of the motion control models were applied for teleoperation of the ISRs along flexible pathways.These include teleoperation models based on three different methods proposed for master-and-slave(MS)tracking.Relationships in terms of configurations of the MS workspace were considered for motion control mapping.Constraints from isometric and non-isometric workspace configurations were modeled to minimize tracking errors and time delays due to underactuation in the flexible robotic systems,during teleoperation.Different studies were carried out to investigate teleoperation of both ISRs along flexible anatomical pathways.Preliminary results show that the MS tracking models can achieve precise and safe motion control of the surgical robotic systems. |
PDF Full Text Request