.png){kind=icon}
.png){kind=icon}
Return
Data-driven velocity control for over-ground body weight support in gait rehabilitation
.png){kind=icon}
PRE
AI
T
V
Y
DOI：10.1016/j.engappai.2026.115723.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper proposes a data-driven velocity control framework for an over-ground body weight support system to enhance gait rehabilitation. Surface electromyography signals are employed to capture neuromuscular activity and infer user intent. By extracting multi-domain features and transforming them into time-frequency representations through continuous wavelet transforms, a convolutional neural network is trained to predict walking velocity with an accuracy of 87 percent. The predicted velocity serves as the reference for a proportional-integral-derivative controller whose gains are automatically optimized using a genetic algorithm, enabling adaptive and stable wheel actuation. Simulation results demonstrate that the integration of data-driven intent recognition and evolutionary control tuning yields precise trajectory tracking with average velocity errors below 0.03 m per second and positional errors under 30 mm. The proposed method exemplifies how artificial intelligence and data-driven control strategies can advance rehabilitation robotics by providing personalized, real-time assistance aligned with natural human movement. This framework shows promise for intelligent rehabilitation systems capable of improving mobility, safety, and engagement during gait training.
Journal
IF:
8
Papers:
5.2K
Citations:
3.5W