A Fuzzy Reinforcement Learning Approach for Continuum Robot Control

Continuum robots (CRs) hold great potential for many medical and industrial applications where compliant interaction within the potentially confined environment is required. However, the navigation of CRs poses several challenges due to their limited actuation channels and the hyper-flexibility of t...

Full description

Saved in:
Bibliographic Details
Published in:Journal of intelligent & robotic systems Vol. 100; no. 3-4; pp. 809 - 826
Main Authors: Goharimanesh, M., Mehrkish, A., Janabi-Sharifi, F.
Format: Journal Article
Language:English
Published: Dordrecht Springer Netherlands 01.12.2020
Springer
Springer Nature B.V
Subjects:
Actuation
Algorithms
Analysis
Artificial Intelligence
Confined spaces
Control
Electrical Engineering
Engineering
Evolutionary algorithms
Game theory
Genetic algorithms
Industrial applications
Learning
Machine learning
Mechanical Engineering
Mechatronics
Robot control
Robotics
Robots
Taguchi methods
Tracking
Fuzzy control
ontinuum robot
Evolutionary algorithms
Reinforcement learning
Taguchi method
ISSN:0921-0296, 1573-0409
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Continuum robots (CRs) hold great potential for many medical and industrial applications where compliant interaction within the potentially confined environment is required. However, the navigation of CRs poses several challenges due to their limited actuation channels and the hyper-flexibility of their structure. Environmental uncertainty and characteristic hysteresis in such procedures add to the complexity of their operation. Therefore, the quality of trajectory tracking for continuum robots plays an essential role in the success of the application procedures. While there are a few different actuation configurations available for CRs, the focus of this paper will be placed on tendon-driven manipulators. In this research, a new fuzzy reinforcement learning (FRL) approach is introduced. The proposed FRL-based control parameters are tuned by the Taguchi method and evolutionary genetic algorithm (GA) to provide faster convergence to the Nash Equilibrium. The approach is verified through a comprehensive set of simulations using a Cosserat rod model. The results show a steady and accurate trajectory tracking capability for a CR.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0921-0296
1573-0409
DOI:10.1007/s10846-020-01237-6