Evaluating the Performance of Joint Angle Estimation Algorithms on an Exoskeleton Mock-Up via a Modular Testing...

Area of Science:

• Robotics
• Biomechanics
• Machine Learning

Background:

• Exoskeleton control requires accurate operator intent detection for seamless actuation.
• Joint angle estimation algorithms typically use multiple sensors, but single-sensor approaches are less explored.
• Operator intent is crucial for effective human-machine system integration in exoskeletons.

Purpose of the Study:

• To evaluate the performance of a kinematic extrapolation algorithm and a Random Forest machine learning algorithm for joint angle estimation using only single-sensor data.
• To assess the feasibility of a modular testing approach for exoskeleton mock-up evaluation.
• To compare the accuracy and actuation time of two distinct joint angle estimation models.

Main Methods:

• A modular testing approach was implemented for exoskeleton mock-up evaluation.
• Two joint angle estimation models, a kinematic extrapolation algorithm and a Random Forest algorithm, were tested.
• Each model was solely informed by kinematic gait data from a single potentiometer on an ankle exoskeleton mock-up.

Main Results:

• The Random Forest algorithm demonstrated lower realized errors in estimated joint angles compared to the kinematic model.
• The Random Forest algorithm resulted in a decreased actuation time.
• The modular testing approach proved feasible for evaluating exoskeleton mock-ups.

Conclusions:

• A single sensor can provide sufficient data for effective joint angle estimation in exoskeleton control.
• The Random Forest machine learning algorithm is a promising approach for single-sensor-based exoskeleton control.
• Modular testing facilitates robust evaluation of human-machine systems in exoskeleton development.