Kinematics Constraint Modeling for Flexible Robots based on Deep Learning 1

Application of flexible robotic systems and teleoperated control recently used in minimally invasive surgery have introduced paradigm shift in interventional surgery. While Prototypes of flexible robots have been proposed for surgical diagnostic and treatments, precise constraint control models are...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference Ročník 2021; s. 4940
Hlavní autoři: Omisore, Olatunji Mumini, Wang, Lei
Médium: Journal Article
Jazyk:angličtina
Vydáno: United States 01.11.2021
Témata:
Biomechanical Phenomena
Deep Learning
Minimally Invasive Surgical Procedures
Robotic Surgical Procedures
Robotics
ISSN:2694-0604
On-line přístup:Zjistit podrobnosti o přístupu
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Application of flexible robotic systems and teleoperated control recently used in minimally invasive surgery have introduced paradigm shift in interventional surgery. While Prototypes of flexible robots have been proposed for surgical diagnostic and treatments, precise constraint control models are still needed for flexible pathway navigation. In this paper, a deep learning based kinematics model is proposed for motion control of flexible robots. Unlike previous approach, this study utilized the different layers of deep learning system for learning the best features to predict the damping value for each point in the robot's workspace. The method uses differential Jacobian to solve IK for given targets. Optimal damping factor that converges precisely around given target is rapidly predicted by a DNN. Simulation of the robot and implementation of the proposed control models are done in V-rep and Python. Validation with arbitrary points shows the deep-learning approach requires an average of 26.50 iterations, a mean error of 0.838, and an execution time of 3.6 ms for IK of single point; and converges faster than other existing methods.
ISSN:2694-0604
DOI:10.1109/EMBC46164.2021.9630418