Publication detail
Artificial Neural Network Application To Walk Of A Four Legged Robot
BEZDÍČEK, M. GREPL, R. ŠVEHLÁK, M. CHMELÍČEK, J.
Czech title
Artificial Neural Network Application To Walk Of A Four Legged Robot
English title
Artificial Neural Network Application To Walk Of A Four Legged Robot
Type
conference paper
Language
cs
Original abstract
This paper presents simple method to extrapolate walk of a four legged robot. Movement of the robot is actuated by twelve servo-drives. This technique of walking is based on a few stable positions of robot body. To reach smooth movement, is sequence of positions extrapolated by artificial neural network. ANN is trained directly on values of servo angles. The servo angles are corresponding to the stable positions. Created neural network is able to extrapolate any posture for continuously changing input parameter. The input parameter is information about required position of the movement. Outputs of network are values of servo angles. Speed of the movement is dependent on the input parameter alternation rate. Introduced technique of walking is useful for plain surface moving.
Czech abstract
This paper presents simple method to extrapolate walk of a four legged robot. Movement of the robot is actuated by twelve servo-drives. This technique of walking is based on a few stable positions of robot body. To reach smooth movement, is sequence of positions extrapolated by artificial neural network. ANN is trained directly on values of servo angles. The servo angles are corresponding to the stable positions. Created neural network is able to extrapolate any posture for continuously changing input parameter. The input parameter is information about required position of the movement. Outputs of network are values of servo angles. Speed of the movement is dependent on the input parameter alternation rate. Introduced technique of walking is useful for plain surface moving.
English abstract
This paper presents simple method to extrapolate walk of a four legged robot. Movement of the robot is actuated by twelve servo-drives. This technique of walking is based on a few stable positions of robot body. To reach smooth movement, is sequence of positions extrapolated by artificial neural network. ANN is trained directly on values of servo angles. The servo angles are corresponding to the stable positions. Created neural network is able to extrapolate any posture for continuously changing input parameter. The input parameter is information about required position of the movement. Outputs of network are values of servo angles. Speed of the movement is dependent on the input parameter alternation rate. Introduced technique of walking is useful for plain surface moving.
Keywords in English
artificial neural network, walking robot, gait
RIV year
2004
Released
01.01.2004
Location
Svratka
ISBN
80-85918-88-9
Book
Engineering mechanics 2004
Pages count
2
BIBTEX
@inproceedings{BUT12234,
author="Milan {Bezdíček} and Robert {Grepl} and Michal {Švehlák} and Jaroslav {Chmelíček},
title="Artificial Neural Network Application To Walk Of A Four Legged Robot",
booktitle="Engineering mechanics 2004",
year="2004",
month="January",
address="Svratka",
isbn="80-85918-88-9"
}