Concentric tube robots are composed of multiple telescopic, concentric, precurved, superelastic tubes that can be axially translated and rotated at their base relative to each other. Kinematic modelling of such tubes is non-trivial due to the interaction of individual tubes with neighboring tubes, to form unique bending curves. Previous research has used conventional iterative Jacobian methods but challenges of determining exact material properties to determine tube interactions remain. We propose the use of deep deterministic policy gradient (DDPG), a deep reinforcement learning algorithm, to learn the inverse kinematics directly. DDPG has been shown to work well in high dimensional continuous space action problems in robotics. DDPG will be evaluated with accuracy criteria for inverse kinematics over different training periods as well as different tube configurations.
A Deep Reinforcement Learning Approach for Inverse Kinematics of Concentric Tube Robots
2019-03-22
In: Proceedings of the 9th Joint Workshop on New Technologies for Computer/Robot Assisted Surgery. (pp. pp. 95-96). CRAS: Genoa, Italy. (2019)
Paper
Elektronische Ressource
Englisch
DDC: | 629 |
Investigating exploration for deep reinforcement learning of concentric tube robot control
BASE | 2020
|General approach for inverse kinematics of nR robots
Online Contents | 2014
|Deep Reinforcement Learning for Concentric Tube Robot Control with a Goal-Based Curriculum
BASE | 2021
|