WAN, Jun ;ZHOU, Zihao ;TANG, Jinlong ;YUN, Nuo ;ZHANG, Xiaoyong ;WANG, Kehong .
Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach.
Articles in Press, [S.l.], v. 0, n.0, p. , october 2025.
ISSN 0039-2480.
Available at: <https://www.sv-jme.eu/article/friction-compensation-and-external-force-estimation-for-robotic-systems-using-a-fuzzy-neural-network-approach/>. Date accessed: 11 feb. 2026.
doi:http://dx.doi.org/.
Wan, J., Zhou, Z., Tang, J., Yun, N., Zhang, X., & Wang, K.
(0).
Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach.
Articles in Press, 0(0), .
doi:http://dx.doi.org/
@article{.,
author = {Jun Wan and Zihao Zhou and Jinlong Tang and Nuo Yun and Xiaoyong Zhang and Kehong Wang},
title = {Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach},
journal = {Articles in Press},
volume = {0},
number = {0},
year = {0},
keywords = {Stribeck model; fuzzy neural network; friction compensation; external force estimation; },
abstract = {To address the issue of inaccurate external force estimation caused by nonlinear friction in high-precision robotic control, this paper proposes a friction compensation and external force estimation method based on the Adaptive Neuro-Fuzzy Inference System (ANFIS). A nonlinear Stribeck friction model is established, combining the Takagi-Sugeno fuzzy system and a neural network for precise identification of joint friction parameters. Compared to traditional Fuzzy Neural Networks (FNN), ANFIS significantly reduces the identification error, particularly at low speeds and near zero-velocity points. A feedforward compensation strategy is designed based on the identified friction model to suppress torque fluctuations caused by friction. To improve external force estimation accuracy, a generalized momentum observer (GM) with friction compensation is developed, and signal processing is performed using median and Butterworth low-pass filters. The median filter removes impulse noise, while the Butterworth filter smooths high-frequency components, avoiding phase distortion. Experiments on a SCARA robot platform, validated with a six-dimensional force sensor, demonstrate the method's effectiveness. Compared to the traditional GM observer, the ANFIS-GM method reduces the Root Mean Square Error (RMSE) and Maximum Absolute Error (MAE) by 18.3% and 27.9%, respectively, and increases the coefficient of determination (R²) to 0.994. The contributions include: (1) the proposed ANFIS-Stribeck hybrid model for nonlinear friction identification; (2) the design of a GM observer with integrated friction compensation; (3) the method's practical value in direct teaching and human-robot interaction for low-cost, high-precision force control.},
issn = {0039-2480}, pages = {}, doi = {},
url = {https://www.sv-jme.eu/article/friction-compensation-and-external-force-estimation-for-robotic-systems-using-a-fuzzy-neural-network-approach/}
}
Wan, J.,Zhou, Z.,Tang, J.,Yun, N.,Zhang, X.,Wang, K.
0 October 0. Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach. Articles in Press. [Online] 0:0
%A Wan, Jun
%A Zhou, Zihao
%A Tang, Jinlong
%A Yun, Nuo
%A Zhang, Xiaoyong
%A Wang, Kehong
%D 0
%T Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach
%B 0
%9 Stribeck model; fuzzy neural network; friction compensation; external force estimation;
%! Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach
%K Stribeck model; fuzzy neural network; friction compensation; external force estimation;
%X To address the issue of inaccurate external force estimation caused by nonlinear friction in high-precision robotic control, this paper proposes a friction compensation and external force estimation method based on the Adaptive Neuro-Fuzzy Inference System (ANFIS). A nonlinear Stribeck friction model is established, combining the Takagi-Sugeno fuzzy system and a neural network for precise identification of joint friction parameters. Compared to traditional Fuzzy Neural Networks (FNN), ANFIS significantly reduces the identification error, particularly at low speeds and near zero-velocity points. A feedforward compensation strategy is designed based on the identified friction model to suppress torque fluctuations caused by friction. To improve external force estimation accuracy, a generalized momentum observer (GM) with friction compensation is developed, and signal processing is performed using median and Butterworth low-pass filters. The median filter removes impulse noise, while the Butterworth filter smooths high-frequency components, avoiding phase distortion. Experiments on a SCARA robot platform, validated with a six-dimensional force sensor, demonstrate the method's effectiveness. Compared to the traditional GM observer, the ANFIS-GM method reduces the Root Mean Square Error (RMSE) and Maximum Absolute Error (MAE) by 18.3% and 27.9%, respectively, and increases the coefficient of determination (R²) to 0.994. The contributions include: (1) the proposed ANFIS-Stribeck hybrid model for nonlinear friction identification; (2) the design of a GM observer with integrated friction compensation; (3) the method's practical value in direct teaching and human-robot interaction for low-cost, high-precision force control.
%U https://www.sv-jme.eu/article/friction-compensation-and-external-force-estimation-for-robotic-systems-using-a-fuzzy-neural-network-approach/
%0 Journal Article
%R
%&
%P 1
%J Articles in Press
%V 0
%N 0
%@ 0039-2480
%8 2025-10-15
%7 2025-10-15
Wan, Jun, Zihao Zhou, Jinlong Tang, Nuo Yun, Xiaoyong Zhang, & Kehong Wang.
"Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach." Articles in Press [Online], 0.0 (0): . Web. 11 Feb. 2026
TY - JOUR
AU - Wan, Jun
AU - Zhou, Zihao
AU - Tang, Jinlong
AU - Yun, Nuo
AU - Zhang, Xiaoyong
AU - Wang, Kehong
PY - 0
TI - Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach
JF - Articles in Press
DO -
KW - Stribeck model; fuzzy neural network; friction compensation; external force estimation;
N2 - To address the issue of inaccurate external force estimation caused by nonlinear friction in high-precision robotic control, this paper proposes a friction compensation and external force estimation method based on the Adaptive Neuro-Fuzzy Inference System (ANFIS). A nonlinear Stribeck friction model is established, combining the Takagi-Sugeno fuzzy system and a neural network for precise identification of joint friction parameters. Compared to traditional Fuzzy Neural Networks (FNN), ANFIS significantly reduces the identification error, particularly at low speeds and near zero-velocity points. A feedforward compensation strategy is designed based on the identified friction model to suppress torque fluctuations caused by friction. To improve external force estimation accuracy, a generalized momentum observer (GM) with friction compensation is developed, and signal processing is performed using median and Butterworth low-pass filters. The median filter removes impulse noise, while the Butterworth filter smooths high-frequency components, avoiding phase distortion. Experiments on a SCARA robot platform, validated with a six-dimensional force sensor, demonstrate the method's effectiveness. Compared to the traditional GM observer, the ANFIS-GM method reduces the Root Mean Square Error (RMSE) and Maximum Absolute Error (MAE) by 18.3% and 27.9%, respectively, and increases the coefficient of determination (R²) to 0.994. The contributions include: (1) the proposed ANFIS-Stribeck hybrid model for nonlinear friction identification; (2) the design of a GM observer with integrated friction compensation; (3) the method's practical value in direct teaching and human-robot interaction for low-cost, high-precision force control.
UR - https://www.sv-jme.eu/article/friction-compensation-and-external-force-estimation-for-robotic-systems-using-a-fuzzy-neural-network-approach/
@article{{}{.},
author = {Wan, J., Zhou, Z., Tang, J., Yun, N., Zhang, X., Wang, K.},
title = {Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach},
journal = {Articles in Press},
volume = {0},
number = {0},
year = {0},
doi = {},
url = {https://www.sv-jme.eu/article/friction-compensation-and-external-force-estimation-for-robotic-systems-using-a-fuzzy-neural-network-approach/}
}
TY - JOUR
AU - Wan, Jun
AU - Zhou, Zihao
AU - Tang, Jinlong
AU - Yun, Nuo
AU - Zhang, Xiaoyong
AU - Wang, Kehong
PY - 2025/10/15
TI - Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach
JF - Articles in Press; Vol 0, No 0 (0): Articles in Press
DO -
KW - Stribeck model, fuzzy neural network, friction compensation, external force estimation,
N2 - To address the issue of inaccurate external force estimation caused by nonlinear friction in high-precision robotic control, this paper proposes a friction compensation and external force estimation method based on the Adaptive Neuro-Fuzzy Inference System (ANFIS). A nonlinear Stribeck friction model is established, combining the Takagi-Sugeno fuzzy system and a neural network for precise identification of joint friction parameters. Compared to traditional Fuzzy Neural Networks (FNN), ANFIS significantly reduces the identification error, particularly at low speeds and near zero-velocity points. A feedforward compensation strategy is designed based on the identified friction model to suppress torque fluctuations caused by friction. To improve external force estimation accuracy, a generalized momentum observer (GM) with friction compensation is developed, and signal processing is performed using median and Butterworth low-pass filters. The median filter removes impulse noise, while the Butterworth filter smooths high-frequency components, avoiding phase distortion. Experiments on a SCARA robot platform, validated with a six-dimensional force sensor, demonstrate the method's effectiveness. Compared to the traditional GM observer, the ANFIS-GM method reduces the Root Mean Square Error (RMSE) and Maximum Absolute Error (MAE) by 18.3% and 27.9%, respectively, and increases the coefficient of determination (R²) to 0.994. The contributions include: (1) the proposed ANFIS-Stribeck hybrid model for nonlinear friction identification; (2) the design of a GM observer with integrated friction compensation; (3) the method's practical value in direct teaching and human-robot interaction for low-cost, high-precision force control.
UR - https://www.sv-jme.eu/article/friction-compensation-and-external-force-estimation-for-robotic-systems-using-a-fuzzy-neural-network-approach/
Wan, Jun, Zhou, Zihao, Tang, Jinlong, Yun, Nuo, Zhang, Xiaoyong, AND Wang, Kehong.
"Friction Compensation and External Force Estimation for Robotic Systems Using a Fuzzy Neural Network Approach" Articles in Press [Online], Volume 0 Number 0 (15 October 2025)
