In reality most of the stuff could be automated with a combination gps / elevation map. Then have some basic object pictures variable sizing etc and you could pretty easily make it completely automated besides some very basic object placement.

The more I think about it the more it could work in train scenarios, to see where and why you do what you do.