Main topics covered:
Lanelets contributions to Autoware.ai. Simon and Ryohsuke have been working hard at landing the last merge requests.
Map status for Autoware.Auto. Brian described the Hackathon and the desire to port the minimal self-driving stack to Autoware.Auto. There is still a huge amount to do. Once the control elements land (PID controller and pure pursuit), work can really start on the path planning elements (mission/global planning, behaviour planning and trajectory generation). We expect that this work will really be done in the Autoware WG and that the requirements on maps will come from that WG to the maps WG.
Example maps in OSM XML Lanelet2 format:
An example map of the AutonomouStuff car park in San Jose can be downloaded here. Please send comments/feedback to @Brian_Holt or @Angelo_Mastroberardi.
Sorry I missed the discussion on coordinate frames. I think Rep 105 is definitely the way to go, but I’ll keep that discussion to the other thread. I’ll be traveling for work next week. But if I have a chance I’ll try to talk with some of the working group members about it in more detail.
Hi @Brian_Holt, I downloaded your example map and found that many lanelets have same left and right way Ids. Are these correct? A lanelet should have different left and right ways to form a closed area right?
Thank you @mitsudome-r, @Brian_Holt.
Commenting on last meeting points, also in case I couldn’t attend the meeting on 24th
1- I will start working on OpenPlanner support for Lanelet2 soon.
2- I uploaded the CARLA videos to Youtube. also may be @Masaya_KATAOKA will be interested in sharing this information with the Simulation group.
a) Introduction: https://youtu.be/Z7PLkcY9Mtk
b) Environment creation: https://youtu.be/hp6NjkfnN_g
@Brian_Holt, @Angelo_Mastroberardi
Thank you for sharing the lanelet xml format.
I have checked it with JOSM, and noticed few things.
As @luqiang21 mentioned, lanelets have the same left and right ways, but I think this was mentioned in the earlier meetings.
Lanelets are not cnnected. In order for the lanelets to be treated as connected lanelets, nodes should be shared at the end of a lanelet and start of following lanelet. Your map has separate nodes for each lanelet, which makes them unconnected.
Parking area has type=area, but the sample map in lanelet2 has type=multipolygon for parking. I believe your parking area is treated as Polygon rather than Area. (See here for difference).
Thank you for the fix.
I have confirmed that bug (3) has been fixed, but bug (3) seems to still exist.
The position of the nodes are correct, but two consecutive lanelets should share the node. (the end node of the previous lanelet should have same id with the starting point of the next lanelet).