Multiprocessed training fixed (previously, there has persisted a bug within the plan manager)
Support for different robot types in training as well as in evaluation (so far: tb3, agv-ota, jackal, ridgeback) - loading simply by specifying model parameter in "start_arena_flatland.launch"
Arena planer should be fixed
Rosnav agent name also specifiable as an argument to the simulation launch script