In control theory (robotics, rockets, and other applications) it's very useful to have a rough physical model of a system.
This is relatively straightforward (but tedious) with rigid body diagrams and kinematic analysis.
I'm imagining a workflow where a user builds something, specifies actuators and sensors, and gets easy to work with text files (Mathematica, Matlab, C/C++/Java/Swift/etc) for the full nonlinear system and linearizations about specific configurations.
The ideal would be something that makes it easy to use experimental data to identify parameters that aren't availiable due to incomplete knowledge of the design.
I'm imagining a loop where you start with a model assembly, generate a first guess at dynamics, measure and tune, and feed that back into the design to help optimize things.
Just being able to export dynamics equations or matrices would be huge and there could be a whole suite of things connected with this.
