Compared to automobiles on roads, trains have less degrees of freedom as they are bound to
railroads. Thus, it should be more straight-forward to let them drive autonomously compared
to automobiles. Several autonomous trains and subways already exist; however they operate
on closed tracks. Typical examples are airport trains, also known as people movers. This
paper sketches the conceptual, technical and legal challenges towards autonomously driving
trains on existing railroads that are freely accessible and thus require reliable obstacle
recognition. We try to generalize the experiences made so far in several large-scale research
projects that aim at automating small, secondary railways. We summarize the results of a
prototypical autonomous train system that we called autoBAHN.