Feedforward Neural Network Assisted Configuration Transition Control of Continuum Surgical Manipulators

Continuum manipulators have been increasingly used in surgical applications recently. A continuum surgical manipulator can be controlled to perform useful tasks when partially inserted into a patient’s cavity, enlarging the workspace of the manipulator. During the insertion and retraction, the inserted portion of the manipulator achieves different configurations, and a configuration transition control framework was proposed based on the resolved motion rate control. However, the inverse kinematics (IK) under the configuration transition framework suffers from failures from time to time due to the reduced motion capability during configuration transition. This paper hence proposes to improve the IK performance of the configuration transition control with the assistance of a feedforward neural network (FNN). During the iteration to solve the IK, the priority and the value of linear and angular velocities are adjusted by the FNN which takes the current and target end effector poses as inputs. By properly adjusting the task priority and the values of the end effector twists, the converging trajectory of the manipulator is regulated to reduce the cases of being trapped in local minima. Numerical simulations were carried out to validate the proposed method. The results show that the proposed method achieves a higher success rate than the original configuration transition method.