Laboratory for Dynamics of Machines and Structures 
Distilling free-form natural laws from experimental data

The conventional way of determining the fundamental equations of motion of dynamical systems requires knowledge of physical laws and the ability to formulate equations by observing the structure and dynamical response of the system. The question is whether we can automatically determine the equation of motion from experimental data.
The task involves the creation of a Python program for the automatic extraction of equations of motion from measurements of the dynamic response of the system using the symbolic regression method within the package gplearn. First, the design of the program is tested on a simple synthetic experiment (physical pendulum) and then on the measurements of a real dynamical system, where the student himself sets up the experiment and performs the measurements.

gplearn package

Contact: Janko Slavič