Related Experiment Video
Updated: May 3, 2026
06:52
An Inertial Measurement Unit Based Method to Estimate Hip and Knee Joint Kinematics in Team Sport Athletes on the Field
Published on: May 26, 2020
8.1K
Machine Learning Enables Rapid Detection of Slips Using a Robotic Hip Exoskeleton
Reese R Peterson1, Jennifer K Leestma2, Inseung Kang3
1Department of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, 30332 USA. He is now with the Florida Institute for Human and Machine Cognition, Pensacola, FL, 32502 USA.
IEEE Transactions on Medical Robotics and Bionics
|August 21, 2025
Summary
Researchers developed slip detection algorithms for hip exoskeletons to aid balance recovery. Extreme Gradient Boosting (XGBoost) showed the best performance, achieving high accuracy and rapid detection times for industrial workers and older adults.
Area of Science:
- Robotics
- Biomechanics
- Machine Learning
Background:
- Slips are a leading cause of injury for industrial workers and older adults.
- Effective balance recovery systems require rapid slip detection.
- Current assistive devices lack effective real-time slip detection capabilities.
Purpose of the Study:
- To compare the efficacy of different machine learning models for detecting slips using hip exoskeleton sensors.
- To identify the optimal algorithm for rapid and accurate slip detection to enhance balance recovery.
Main Methods:
- Trained and evaluated user-independent models using Linear Discriminant Analysis (LDA), Extreme Gradient Boosting (XGBoost), and Convolutional Neural Networks (CNN).
- Utilized native sensor data from a hip exoskeleton during treadmill-based slip perturbations.
- Tested models on early-stance (ES) and late-stance (LS) slips of varying magnitudes.
Main Results:
- All tested models, except LDA for LS slips, achieved over 90% accuracy in slip detection.
- XGBoost demonstrated superior performance with rapid detection times (155.06 ms for ES, 228.88 ms for LS) and high median accuracies (96.25% for ES, 93.75% for LS).
- XGBoost achieved 100% sensitivity at 195.64 ms (ES) and 266.24 ms (LS).
Conclusions:
- Machine learning models, particularly XGBoost, show significant promise for real-time slip detection in hip exoskeletons.
- The developed models can aid in designing effective balance recovery strategies for slip perturbations.
- Further research is warranted to create generalizable models for diverse populations and slip scenarios.
