Closed osrf-migration closed 5 years ago
Original comment by Martin Dlouhy (Bitbucket: robotikacz).
This is probably related to #41 ... what we had to do was to return to Base Station in order to be sure ...
Original comment by Nate Koenig (Bitbucket: Nathan Koenig).
The scoring server is receiving a major upgrade in pull request #117. Scoring will still have to occur through the provided comms model. For how scoring works please refer to the guidelines in https://www.subtchallenge.com/resources/SubT_Challenge_Competition_Rules_Tunnel_Circuit.pdf. That document may answer your other comments.
Original comment by Nate Koenig (Bitbucket: Nathan Koenig).
We are adding an ACK to the base station that is sent back to a caller when they report the position of an artifact. Will an inclusion of an ACK resolve this issue?
Original comment by Alfredo Bencomo (Bitbucket: bencomo).
This has been implemented, but re-open it if there are addition problems or suggestions.
Original report (archived issue) by Jack Kelly (Bitbucket: jack-kelly).
Because the comms is following a simple UDP-like model and we get no delivery guarantees, it seems strange to set up scoring to punish multiple packet submissions. If the base station communications callbacks (or perhaps just the class itself) were exposed in the same way that the robots are, we could implement handshakes or other delivery guarantees before calling a
submitArtifact()
function.Or scoring could be modified to not remove artifacts from the list when a submission is received and instead take the closest each time, possibly replacing an old score on that particular artifact.
Or we could switch to a TCP like comms model with delivery receipts that has rate-throttling instead of the micro level packet dropping modeling that the current model has.