2024-12-16 ODD working group weekly meeting (218) #5553
Igata-ctrl
started this conversation in
Working group meetings
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Attendee
Agenda/Minutes
On November 17, 2024, 295/340 "public road bus" and "pull out" scenarios succeeded on Autoware.universe main branch.
110/142 public road bus scenarios were successful.
183/198 pull out scenarios were successful. 15 scenarios failed.
On November 17, 77/96 Dense Urban scenarios were successful on Autoware.universe scenario-dev branch.
On November 24, 2024, 297/340 "public road bus" and "pull out" scenarios succeeded on Autoware.universe main branch.
112/142 public road bus scenarios were successful.
185/198 pull out scenarios were successful.
On November 24, 75/96 Dense Urban scenarios were successful on Autoware.universe scenario-dev branch.
On December 1, 2024, 299/340 "public road bus" and "pull out" scenarios succeeded on Autoware.universe main branch.
113/142 public road bus scenarios were successful.
186/198 pull out scenarios were successful.
On December 1, 76/96 Dense Urban scenarios were successful on Autoware.universe scenario-dev branch.
Summary of currently filing scenarios.
On December 8, 2024, Autoware build failed and no new result was obtained.
On December 15, 2024, 294/340 "public road bus" and "pull out" scenarios succeeded on Autoware.universe main branch.
112/142 public road bus scenarios were successful.
182/198 pull out scenarios were successful.
On December 15, a build error occurred with "scenario-dev" branch and no result obtained for Dense Urban scenarios (and LSV for Reference Design WG).
Troubleshoot of Huy's on-premise simulator setup. The long path to the map files may have been the cause.
[recap of the discussions until last week]
As for the following scenarios in the Low Speed Vehicle suite (Reference Design WG), both worked fine after deleting Npc1's "Max speed = 0 m/s" by clicking the trash can icon to the right. This may apply to other failing scenarios in the LSV suite.
Continued discussion of the Jay walkers use case including the vehicle-pedestrian interface, the (pedestrian) intention estimation, etc.
As for the new scenario suite (catalog) for LSV, it seems that almost all the scenario files are just changed with names. May need to ask Daniel-san to revert all the titles to the original one and take the following time-consuming process to duplicate and modify the copied scenarios for LSV/Reference Design use.
Started discussion of the Jay walkers use case.
Had a Safety Pool/AWF meeting regarding the existing issues.
Discussed with Daniel whether the scenarios are appropriately copied to LSV catalog instead of modifying the original. -> @Igata-ctrl to double check it by next week.
Preliminary discussion with TIER IV Evaluator team revealed that the import of OpenSCENARIO 1.x files will be supported hopefully by the end of 2024. Currently, Scenario Simulator v2 uses the proprietary t4v2 format in yaml to better handle the parameter variations. This nee import function enables the use of out side scenario editor like AVL SCENIUS™️ Scenario Designer and help importing scenarios from Safety Pool database, though the issue of map format and the lack of destinations may remain.
Discussion to create Evaluator improvement wish list.
Discussed the improvement requests for Evaluator.
Regarding the needs for calculating the accurate RSS safety distance from Vf(Npc) and Vr(Ego), TIER IV Mapping and Localization team will be able to add the functionality using
UserDefinedValueCondition
like UC-v2-F-15-01001_001_case01_cmn_general case. It may take more than two weeks depending on their resource margin/availability. -> @brkay54 to kindly submit an issue regarding this request mentioning Takeuchi-san of TIER IV (@kosuke55) who is working onUserDefinedValueCondition
.For the current scenario suite for weekly execution above, the RSS definition for minimum safety distance and 2.499m/s^2 for maximum comfortable deceleration have been applied.
@Igata-ctrl got the information in Simulation working group meeting that the resource allocated to Scenario Simulation v2 on the cloud may not be sufficient and it can make the simulation result not trustworthy enough. We may need to analyze the Simulation Log file to see if there are many waring messages in there when we get unexpected simulation result.
@Igata-ctrl to investigate if we can use real time speed of Ego and Npc to calculate RSS Safe Distance more accurately.
Discussion (RSS distance and comfortable deceleration) pages have been added to the Use cases for Dense Urban Area 2024 presentation.
Reached a consensus to use RSS definition for minimum safety distance and 2.499m/s^2 for maximum comfortable deceleration for low speed shuttle.
Discussion of the condition/threshold setting so that we can better track the Autoware maturity through scenario simulation.
After the discussion, it turned out "min_acc" parameter should better be reverted to original value of -1.5 (this week).
Current discussion with Safety Pool
Action Item
Documents
The bus ODD use case list which was assigned to the members is here.
The bus ODD use case list has been also uploaded here to make it visible to anyone on the net.
The ODD working group shared document folder is here. [Restored!]
Discussions and Q&As in AWF Discord ODD WG channel are also encouraged.
Tools
[Autoware Evaluator (CI/CD pipeline)]
[Scenario Simulator V2 (Scenario testing framework)]
The GUI Scenario editor
[TIER IV account]
Administrative
The two meetings on December 25th (Christmas day) and January 1st (New Year day) have been cancelled.
Please check the ODD WG wiki page.
Recurring weekly meetings have been scheduled. Please check Autoware Foundation events calendar and add this calendar to your own Google calendar by clicking the right bottom button and/or add your contact to ODD WG invitation group to receive invitations for future meetings.
ODD WG meetings are held weekly in the following single time slot.
Beta Was this translation helpful? Give feedback.
All reactions