A Motion Intention Recognition Method for Lower-Limb Exoskeleton Assistance in Ultra-High-Voltage Transmission Tower...

Area of Science:

• Robotics and Human-Computer Interaction
• Biomedical Engineering
• Artificial Intelligence

Background:

• Transmission tower climbing is essential for ultra-high-voltage power and communication infrastructure maintenance.
• Existing lower-limb exoskeletons face challenges in accurately recognizing user motion intentions in complex environments.

Purpose of the Study:

• To develop an improved motion intention recognition method for lower-limb exoskeletons used in transmission tower climbing.
• To enhance the accuracy and reliability of exoskeleton control in demanding operational settings.

Main Methods:

• Utilized inertial measurement unit (IMU)-based bidirectional temporal deep learning.
• Employed a one-dimensional convolutional neural network (1D-CNN) for local feature extraction.
• Integrated a bidirectional long short-term memory network (Bi-LSTM) with a temporal attention mechanism for sequence modeling.

Main Results:

• The proposed deep learning method significantly outperformed traditional machine learning and unidirectional models in accuracy and F1-score.
• Demonstrated superior performance in identifying flexion/extension phases and transitional movement states.
• Achieved precise recognition of short-duration and transitional motions critical for climbing.

Conclusions:

• The developed IMU-based bidirectional temporal deep learning method offers a robust solution for motion intention recognition in lower-limb exoskeleton control for tower climbing.
• Provides a valuable offline analysis tool and a reference for developing advanced assistive control strategies for power maintenance tasks.