[GO Feedback] Score for manipulation items feel too low #950
Comments
LeroyR
changed the title
|
In theory, no team can score any points (including bonuses) unless they perform a main goal task. Also, I am against too permissive Deus ex Machina, at least for core skills in a given task (for example, bypass manipulation in a manipulation task or perception in a perception task). |
|
"optional goal "one time a full transporting without human help extra points" |
LeroyR
added a commit
that referenced
this issue
Apr 15, 2025
** Note: Your contribution is expected to meet the conventions and policies described in the [contribution guidelines](https://github.com/RoboCupAtHome/RuleBook/wiki/Guidelines:-Contributing) ** ## Description Closes issue #950 The Manipulation tasks (storing groceries, cleanup) are focused on repetition even though the manipulation rate of most teams is pretty small. With the consolation points for navigation and recognition the difference between a team having grasped and placed something and the team only recognizing and showing bounding boxes can be _negative_: Team a: nav + 5 bounding boxes: 90pts Team b: nav + pick (wrong detection) + place (wrong location): 80pts With deus ex machina this can increase even more. Due the repetitive nature a increase in the points for pick/place overvalues the individual task thus i opted for just adding a one time bonus. This change: - adds two 50 point (one time) boni for pick and place. - Adds penalty for wrong detection (e.g showing 7 detection but 4 of them are wrong should score 3*15) ## Additional Remarks - Is this our preferred solution? - Should something like this also be added to cleanup?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Even if the max score of the receptionist vs e.g. storing groceries is comparable. The difference between a Team doing a single pick and place and a team doing no manipulation is just 100 points.
I think the 'reduced points for repetition' as in egpsr may be a good solution allowing fast/stable/reliable robots to continue scoring while increasing the difference between at least one successful pick and place and none at all. Increasing the difference between some and no manipulation while not overvaluing manipulation focussed tasks. Currently all manipulation tasks are repetition heavy and a blanket increase would inflate the scores.
The text was updated successfully, but these errors were encountered: