Abstract:Introduction Musculoskeletal modeling (MSKm) and simulation provide an ideal framework for estimating movement-related parameters such as joint contact forces which cannot be measured in vivo. However, these parameters are not well-documented in free-living situations as inertial measurement units (IMUs) have rarely been combined with MSKm workflows. Although IMU-based systems are valuable alternatives to camera-based systems to estimate kinematics, no robust and validated method exists to estimate ground reaction forces (GRF) thereby alleviating the need for force plates. The aim of this study was to implement and validate a workflow for GRF estimation using a combined IMU and MSKm workflow. The ability to estimate such information, when combined with MSKm would provide clinicians access to movement-related relevant to the onset/progression of musculoskeletal diseases in an ecological environment. Research question Can IMU-based kinematics combined with an MSKm-based simulation workflow be used to accurately estimate GRF during gait? Methods Two healthy subjects (1 man, 1 female; age: 27.5±2.2y) walked overground (self-selected speed). MotionCapture (100Hz) and InertialCapture (60Hz) systems were used to record 3D-marker trajectories and body segment orientations. GRF were recorded using force plates (1000Hz). MoCap and InCap kinematics were calculated using OpenSim [1] and sensors-to-segment calibration OpenIMUs [2], respectively. Musculoskeletal model [3] containing 35 DOFs, 49 muscles, and 5 foot contact spheres was used in both approaches. OpenSim Moco was used to track the IMU-based kinematics with two main goals: (i) minimize the error between input kinematics and simulated motion, (ii) minimize the sum of the absolute value of the muscle activations. MocoSolver was applied to compute the optimized motion and muscle activations. GRF were evaluated based on the generated forces of the foot contact spheres. Centre of pressure (COP) was estimated using a Zero Moment Point (ZMP) method [4]. Results RMSE of the joint angle between MoCap and InCap was less 5° for all joint angles. RMSE between InCap kinematics and Moco simulated kinematics were 1-6° for all joint angles with knee and ankle flexion/extension max error of 10°. RMSE of anterior/posterior, vertical, and medio/lateral GRF was 0.10±0.02 BW, 0.25±0.01 BW and 0.06±0.03 BW respectively, compared to measured force plate data (Fig. 1). RMSE of anterior/posterior, and medio/lateral COP was 6.7±0.9cm and 3.8±0.4cm respectively, compared to measured force plate data. Discussion The developed workflow for the estimation of GRF based on IMU-based kinematics and OpenSim Moco provided promising results. However, the contact sphere definition strongly affects the estimated GRF with a direct effect on the calculated COP and ultimately joint moments (not shown).

Estimation of Normal Ground Reaction Forces in Multiple Treadmill Skiing Movements Using IMU Sensors with Optimized Locations

Estimation of alpine skier posture using machine learning techniques

Influence of IMU position and orientation placement errors on ground reaction force estimation

Biodynamic Analysis of Alpine Skiing with a Skier-Ski-Snow Interaction Model

Estimating Ground Reaction Forces from Inertial Sensors

A Ground Reaction Force Model of Speed Skating Based on Non-Contact Measurement System

IMU-based ground reaction force estimation using OpenSim Moco

Estimation of Mechanical Power Output Employing Deep Learning on Inertial Measurement Data in Roller Ski Skating

Estimation of Ground Reaction Forces during Sports Movements by Sensor Fusion from Inertial Measurement Units with 3D Forward Dynamics Model

SnowMotion: A Wearable Sensor-Based Mobile Platform for Alpine Skiing Technique Assistance

A Method of Attitude Measurement and Level Assessment for Skiers Based on Wearable Inertial Measurement

Estimation of Three-Dimensional Lower Limb Kinetics Data during Walking Using Machine Learning from a Single IMU Attached to the Sacrum

An IMU-Based Ground Reaction Force Estimation Method and Its Application in Walking Balance Assessment.

Ski Position during the Flight and Landing Preparation Phases in Ski Jumping Detected with Inertial Sensors

Algorithm for Gait Parameters Estimation Based on Heel-Mounted Inertial Sensors

3D gait analysis in children using wearable sensors: feasibility of predicting joint kinematics and kinetics with personalized machine learning models and inertial measurement units

Three-Rigid-Body-Particle Modeling and Optimization of Trajectory and Posture for Alpine Skiing

Distributed IMU Sensors for In-Field Dynamic Measurements on an Alpine Ski

3D Knee and Hip Angle Estimation With Reduced Wearable IMUs via Transfer Learning During Yoga, Golf, Swimming, Badminton, and Dance

The Quantitative Prediction of Auxiliary Sliding Distance of Freestyle Skiing Based on MLP Neural Network