Edge Computing for Environment-Based Locomotion Modes Prediction and Terrain Features Calculation in Lower Limb...

Area of Science:

• Robotics and Human-Computer Interaction
• Biomedical Engineering
• Sensor Fusion

Background:

• Wearable lower-limb prostheses require environment-aware control for optimal function.
• On-device computational and size constraints limit real-time perception and decision-making in current prostheses.

Purpose of the Study:

• To develop an edge-computing framework for prosthetic limbs enabling real-time locomotion mode prediction and terrain feature calculation.
• To address the challenges of integrating perception and control within strict on-device limitations.

Main Methods:

• Integration of two inertial measurement units (IMUs) with an event-triggered array LiDAR sensor.
• Dynamic triggering of LiDAR scans based on joint kinematics from IMUs during critical gait phases.
• Deployment of the complete perception-action pipeline on an embedded platform for deterministic real-time operation.

Main Results:

• Achieved an overall locomotion-mode prediction accuracy of 97.85% (98.80% indoor, 96.90% outdoor).
• Calculated terrain features with a mean relative error of 4.87%.
• Maintained an end-to-end decision latency under 11.5 ms.

Conclusions:

• The proposed edge-computing framework effectively enables environment-aware control for lower-limb prostheses under strict constraints.
• The system's high accuracy and low latency support real-time prosthetic control, enhancing user mobility and safety.