Estimating Running Ground Reaction Forces from Plantar Pressure during Graded Running

Area of Science:

• Biomechanics
• Wearable Technology
• Sports Science

Background:

• Ground reaction forces (GRFs) are crucial for understanding runner-environment interactions and inverse dynamics.
• Existing wearable technology primarily predicts GRFs during level-ground running.
• Predicting GRFs across diverse running conditions (speed, incline, decline) remains a challenge.

Purpose of the Study:

• To predict vertical and anterior-posterior GRFs during running across various speeds and slopes.
• To evaluate the efficacy of recurrent neural networks (RNNs) and linear models for GRF prediction.
• To assess the impact of excluding speed and slope as input parameters on prediction accuracy.

Main Methods:

• Eighteen subjects ran on an instrumented treadmill at different speeds and inclines/declines.
• Ground reaction forces (GRFs) and plantar pressure data were collected.
• GRFs were estimated using linear models and an RNN, with speed, slope, and plantar pressure as inputs.

Main Results:

• The RNN model demonstrated superior performance compared to the linear model for all conditions, particularly for anterior-posterior GRFs.
• Excluding speed and slope as input features had minimal impact on prediction accuracy.
• Subject-specific model training significantly reduced prediction errors from 8% to 3%.

Conclusions:

• Wearable-based GRF prediction models, especially RNNs, can accurately estimate forces during varied running conditions.
• The developed models enable researchers to analyze running biomechanics and injury mechanisms in real-world settings.
• Accurate, wearable-based GRF data can advance the understanding of running performance and injury prevention.