Simplest is 1 "program", a bunch of songs in order. Possible UI actions: play program in order from the beginning, go forward/backward one song, repeat program, repeat one song, shuffle program.
This was the model for the iPod shuffle that did not have a display. It is similar to physical media formats: tape, records, CDs (though the latter 2 do have rudimentary random access).
Next is multiple programs: play the default (first) program, go forward/backward one program, repeat all programs, repeat one program, repeat one song. The same song could be included on multiple programs; the device should deduplicate. A program could be an album or playlist.
Is two levels or hierarchy too complicated to navigate without a display? More levels?
Filesystem features required: symbolic links for deduplication, ordering of files and directories, metadata.
Inspiration was the lack of a standardized interface to play a music library stored on an external device on a car stereo, using an interface built into the car.