The idea
Film real marble runs, fit a physics model to measured trials, then rewind a digital copy to test a new ramp or friction setting.
How it works
- Record marble runs with measured ramp dimensions and known objects.
- Fit a simple model to several calibration trials and preserve uncertainty.
- Change friction, slope, or gravity in the digital version, then test a suggested physical adjustment.
AI’s role: Help label objects and suggest a next controlled trial; the simulator and measured parameters calculate outcomes, which still need comparison with physical results.
First demonstration
Record three runs, adjust the model, and test its next physical suggestion. Compare the predicted marble path with the measured one.
What to solve next
How can the model distinguish friction changes from camera measurement error or inconsistent release force?