Pengfeng Li

Pengfeng Li

Auto-generated profile, based on publicly available information

Pengfeng Li

Sichuan University, School of Physics, China

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 August 2012
Verified
Conference Paper
Closed access
JHJin Hu
Jin Hu
ZHZeng-Guang Hou
Zeng-Guang Hou
FZFeng Zhang
Feng Zhang
YCYixiong Chen
Yixiong Chen
PLPengfeng Li
Pengfeng Li
This paper proposes three training strategies based on impedance control, including passive training, damping-active training and spring-active training, for a 3-DOF lower limb rehabilitation robot designed for patients with paraplegia or hemiplegia. Controllers with similar structure are developed for these training strategies, consisting of dual closed loops, the outer impedance control loop and the inner position/velocity control loop, known as position-based impedance control method. Simulation results verify that position-based impedance control approach is feasible to accomplish the training strategies.

7 July 2023
Verified
Conference Paper
Open access
JZJunYue Zhuang
JunYue Zhuang
PLPengfeng Li
Pengfeng Li
MHMingmin Huang
Mingmin Huang
This article proposes a machine learning prediction algorithm that removes subjective feature sequence dimension selection. Under the conditions of imbalanced data distribution, multiple classification categories, and strong subjectivity in user ratings in the user satisfaction scoring dataset provided by Beijing Mobile, a rigorous and reliable optimization algorithm for predicting user satisfaction scoring is developed by integrating Simulated Annealing (SA), Random Forest (RF), and Stacking under known conditions. Based on the characteristic of too many feature sequences in the dataset, we use the Simulated Annealing to search for the feature sequence that could maximize the macro-F1 score of the Random Forest model and construct the prediction dataset with the best performance for the Random Forest model. We also use multiple rounds of grid search to select the optimal Random Forest model parameters. Stacking is used to predict specific scores in each data set. Finally, the constructed scoring prediction model significantly improves prediction accuracy. This algorithm improves the accuracy of scoring prediction from the original 0.55 to approximately 0.7. However, its limitation lies in the fact that it performs better and achieves higher accuracy only in a large sample dataset.