Adaptive Force and Position Control of Magnetic Endoscopes using Reinforcement Learning - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2023

Adaptive Force and Position Control of Magnetic Endoscopes using Reinforcement Learning

Antonio Marino

Résumé

In this work, an adaptive control framework based on model-free reinforcement learning is applied to a magnetically manipulated endoscope to simultaneously control the endoscope position and adapt to the highly variable, dynamic, unstructured environment of the human gastrointestinal tract. The ability to adapt control parameters enables the endoscope to overcome obstacles and successfully navigate the tract in the absence of a motion planning algorithm. This also facilitates maintenance of optimal contact with the tissues, which is beneficial for multiple diagnostic applications-including micro-ultrasound imaging. The approach is experimentally validated using the Magnetic Flexible Endoscope in a benchtop colon simulator through (a) execution of forward and backward motion with contact force control, (b) effectiveness at navigating a sharp turn, and (c) successfully navigating an unmodelled obstacle. Overall, this is the first example of model-free adaptive control being successfully applied to magnetic manipulation and lays the groundwork for the further development of advanced motion planning and autonomous navigation algorithms.
Fichier principal
Vignette du fichier
main.pdf (13.89 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04162017 , version 1 (13-07-2023)

Licence

Domaine public

Identifiants

  • HAL Id : hal-04162017 , version 1

Citer

Antonio Marino. Adaptive Force and Position Control of Magnetic Endoscopes using Reinforcement Learning. 2023. ⟨hal-04162017⟩
154 Consultations
34 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More