Datasets:
Give Town10HD_Opt's lane boundaries the markings CARLA draws
The lanelet2 boundaries had no type, so lanelet2's traffic rules forbade every lane change. The routing graph had adjacentLeft/adjacentRight but no left/right relations, and Autoware found no lane change target.
Each boundary of every lanelet outside a junction now carries the marking CARLA draws there. It is read from the CARLA 0.10 waypoints at a quarter, half and three quarters of the lanelet by tag_lane_boundaries.py (https://gist.github.com/hakuturu583/f11582b8ca65a5e2efae9fe884e1dfc0):
- Broken -> line_thin/dashed (38 boundaries);
- Solid -> line_thin/solid (48);
- SolidSolid -> line_thin/solid_solid (36).
Junction boundaries, and road edges CARLA draws no line on, stay untyped. Where readings disagree, the more restrictive wins.
So lane changes are legal exactly where CARLA paints a broken line, on 38 lanelets each way, and nowhere else. This is not lane changes legal everywhere, which the README warned makes the behavior planner spin.
Checked with autoware_carla_scenario's logical scenarios on CARLA 0.10 with this map, SUMO traffic backend:
- lane_change left/right: 12/12 each (0 cases before);
- lane_change_fail left/right: 26/26 and 40/40;
- cut_in left/right: 7/8 each.
The file is edited as text: two tag lines per typed boundary, nothing else changed.