Hi All!
We had the following question from BART LAB rescue that I thought was worth splitting into its own thread for discussion.
We disagree with “4x multiplier for autonomy”. We believe that in any real situations, humans should also be able to make crucial decisions for the best outcome - in case of an emergency (e.g. further destructions within the building). Therefore, we believe that “autonomy” should be optional. We suggest that “autonomy” can be its own new task, instead of being integrated into every task for extra points, which we believe is way too much.
Of course we are tuning the multipliers so this is a good discussion to have.
Note that the 4x multiplier is for full autonomy, which is already very difficult for most robots. We can have some discussion as to how this applies to “very easy” bays where perhaps it is too much.
For terrains, obstacles, and dexterity, I would suggest that a teleoperated robot being 4x faster than an autonomous robot going through, say, the K-rail terrain on crossover slope, or inspecting all of the pipes (assuming it’s not pre-programmed to fix setpoints and must find the pipes itself), is not unreasonable.
(I’ll let the committee members who are focused on autonomy comment further.)
I will also note that really the interesting change here is the 2x multiplier for what was semi-autonomy and is now for radio degradation. This is the multiplier that is relevant to your scenario whereby the humans make the critical decisions and the robot makes the other decisions.
What we do want to get away from is the situation that robots could compete with the assumption that the communications will be perfect, just like 10 years ago we got away from the situation where robots could compete with the assumption that the floor would be reasonably flat.
Now those are the justifications for having some multiplier but of course the details of the implementation and magnitude of the multipliers is still up for discussion! I can certainly see a reason for changing the values or perhaps having them relevant to only some tests and not others.
What does everyone else think?