Multi-Joint Leg Moment Estimation During Walking Using Thigh or Shank Angles.

Area of Science:

• Biomechanics
• Robotics
• Biomedical Engineering

Background:

• Coordinating robotic prosthetics, orthotics, or exoskeletons with biological limbs is crucial for human-like locomotion.
• Understanding the relationships between lower extremity joint kinematics and kinetics is key to achieving seamless integration.
• Existing methods may require complex switching rules or intermediate parameters for motion coordination.

Purpose of the Study:

• To develop a method for continuously estimating sagittal plane ankle, knee, and hip moments.
• To utilize readily available limb angle data (shank or thigh) as input for moment estimation.
• To circumvent the need for switching rules or intermediate parameters in motion coordination.

Main Methods:

• Employed a nonlinear autoregressive model with exogenous inputs (NARX).
• Integrated neural networks and wavelets for signal processing and estimation.
• Evaluated estimation performance using four case studies with varying input-output combinations and three databases (106 subjects).

Main Results:

• The best estimation performance was achieved when estimating ankle moments from shank angles.
• Estimation accuracy varied with joint, limb segment used for input, and walking speed (e.g., knee moment estimation was less accurate at slow speeds).
• Average RMS errors were within 0.09-0.15 Nm/kg, MAEs within 0.06-0.11 Nm/kg, and correlation coefficients ranged from 0.71 to 0.99, comparable to existing studies.

Conclusions:

• Continuous estimation of lower extremity joint moments from limb angles is feasible using the proposed NARX model.
• Shank angles provide a more accurate input for estimating ankle moments compared to thigh angles for hip and knee moments.
• The developed method offers a promising, parameter-light approach for enhancing control strategies in lower limb robotic devices.