Back to Home

Posture Prediction in Static Pushing/Pulling

Early‑stage research · Force measurement → OpenSim features → ANN posture estimation

Summary: This project builds a pipeline from force measurement to posture prediction. We’re currently designing and fabricating the measurement setup and data‑logging system.
Designing rig Strain gauges & DAQ OpenSim feature extraction ANN (planned)

Project Context & Goal

  • Why: Predicting posture under static pushing/pulling can inform ergonomic design and reduce musculoskeletal risk.
  • Objective: Train an artificial neural network that maps measured static forces to joint postures estimated with OpenSim.
  • Scope: Hardware setup → data logging → biomechanical feature extraction → ANN training/validation.

Process Roadmap

Now

1) Force Measurement Setup

CAD will be added when ready
  • Design & manufacture the mechanical rig for pushing/pulling.
  • Integrate strain gauges; select DAQ; amplifier wiring & calibration.
  • Static load cases & calibration curve generation.

2) Data Collection

Add test case photos / screenshots
  • Implement robust data logging (timestamping, units, metadata).
  • Define test cases; record steady static forces for each posture.
  • Optional: synchronize with motion capture or reference posture measures.

3) Biomechanical Analysis (OpenSim)

Add OpenSim screenshots / figures
  • Process trials in OpenSim; extract joint angles & posture descriptors.
  • Assemble a clean dataset (features/labels) for model training.
  • Train/validation split and normalization strategy.
Next

4) ANN Model Development

Add model diagram & results later
  • ANN architecture selection; regularization & early stopping.
  • Performance metrics (MAE for joint angles, posture error heatmaps).
  • Generalization checks on unseen test cases.