Learning for Skill Refinement in Robotic Systems
Summary :
Learning control is a new approach to the problem of skill refinement for robotic systems by repetitive training. A class of simple learning control algorithms with a forgetting factor and without use of the derivative of velocity signals for motion control of robot manipulators is proposed and the convergence property is discussed. The robustness of such a learning control scheme with respect to initialization errors, disturbances, and measurement noise is studied extensively. It is proved that motion trajectories converge to a neighborhood of the desired one and eventually remain in it. In the argument the passivity of robot dynamics and displacement robot dynamics plays a fundamental role. Relations of the size of attraction neighborhoods with the magnitudes of initialization errors and other disturbances are obtained, which suggests a rule for selection of the forgetting factor in the progress of learning. Based on these results, two classes of learning control called "interval training" and "selective learning" are proposed in order to accelerate the speed of convergence.
- Publication
- IEICE TRANSACTIONS on Fundamentals Vol.E74-A No.2 pp.235-243
- Publication Date
- 1991/02/25
- Publicized
- Online ISSN
- DOI
- Type of Manuscript
- INVITED PAPER
- Category
Authors
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Suguru ARIMOTO, "Learning for Skill Refinement in Robotic Systems" in IEICE TRANSACTIONS on Fundamentals,
vol. E74-A, no. 2, pp. 235-243, February 1991, doi: .
Abstract: Learning control is a new approach to the problem of skill refinement for robotic systems by repetitive training. A class of simple learning control algorithms with a forgetting factor and without use of the derivative of velocity signals for motion control of robot manipulators is proposed and the convergence property is discussed. The robustness of such a learning control scheme with respect to initialization errors, disturbances, and measurement noise is studied extensively. It is proved that motion trajectories converge to a neighborhood of the desired one and eventually remain in it. In the argument the passivity of robot dynamics and displacement robot dynamics plays a fundamental role. Relations of the size of attraction neighborhoods with the magnitudes of initialization errors and other disturbances are obtained, which suggests a rule for selection of the forgetting factor in the progress of learning. Based on these results, two classes of learning control called "interval training" and "selective learning" are proposed in order to accelerate the speed of convergence.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/e74-a_2_235/_p
Copy
@ARTICLE{e74-a_2_235,
author={Suguru ARIMOTO, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Learning for Skill Refinement in Robotic Systems},
year={1991},
volume={E74-A},
number={2},
pages={235-243},
abstract={Learning control is a new approach to the problem of skill refinement for robotic systems by repetitive training. A class of simple learning control algorithms with a forgetting factor and without use of the derivative of velocity signals for motion control of robot manipulators is proposed and the convergence property is discussed. The robustness of such a learning control scheme with respect to initialization errors, disturbances, and measurement noise is studied extensively. It is proved that motion trajectories converge to a neighborhood of the desired one and eventually remain in it. In the argument the passivity of robot dynamics and displacement robot dynamics plays a fundamental role. Relations of the size of attraction neighborhoods with the magnitudes of initialization errors and other disturbances are obtained, which suggests a rule for selection of the forgetting factor in the progress of learning. Based on these results, two classes of learning control called "interval training" and "selective learning" are proposed in order to accelerate the speed of convergence.},
keywords={},
doi={},
ISSN={},
month={February},}
Copy
TY - JOUR
TI - Learning for Skill Refinement in Robotic Systems
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 235
EP - 243
AU - Suguru ARIMOTO
PY - 1991
DO -
JO - IEICE TRANSACTIONS on Fundamentals
SN -
VL - E74-A
IS - 2
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - February 1991
AB - Learning control is a new approach to the problem of skill refinement for robotic systems by repetitive training. A class of simple learning control algorithms with a forgetting factor and without use of the derivative of velocity signals for motion control of robot manipulators is proposed and the convergence property is discussed. The robustness of such a learning control scheme with respect to initialization errors, disturbances, and measurement noise is studied extensively. It is proved that motion trajectories converge to a neighborhood of the desired one and eventually remain in it. In the argument the passivity of robot dynamics and displacement robot dynamics plays a fundamental role. Relations of the size of attraction neighborhoods with the magnitudes of initialization errors and other disturbances are obtained, which suggests a rule for selection of the forgetting factor in the progress of learning. Based on these results, two classes of learning control called "interval training" and "selective learning" are proposed in order to accelerate the speed of convergence.
ER -
English
