Datasets:
Commit ·
f67179a
1
Parent(s): 5058c66
Share Town10HD_Opt's signal heads and merge points its bounds draw twice (#7)
Browse files- Share Town10HD_Opt's signal heads and merge points its bounds draw twice (21fe8ac949ebc0432a93d9904e0504114c6d4573)
- README.md +4 -4
- autoware_maps/Town10HD_Opt/lanelet2_map.osm +2 -2
README.md
CHANGED
|
@@ -46,7 +46,7 @@ CARLA's [Town 10](https://carla-ue5.readthedocs.io/en/latest/map_town10/): an in
|
|
| 46 |
| `lanelet2_map.osm` | `pointcloud_map.pcd` |
|
| 47 |
| --- | --- |
|
| 48 |
| [](https://hakuturu583.github.io/simple_lanelet2/embed.html?map=https%3A%2F%2Fhuggingface.co%2Fdatasets%2FAutowareFoundation%2Fcarla-ue5-maps%2Fresolve%2Fmain%2Fautoware_maps%2FTown10HD_Opt%2Flanelet2_map.osm&theme=dark) |  |
|
| 49 |
-
| 168 lanelets,
|
| 50 |
|
| 51 |
The two are the same town in the two forms Autoware reads it in, drawn to the same frame. What the point cloud shows and the lanelet2 map does not is the reason the point cloud is here at all: this is Unreal Engine 5 geometry, not the 0.9 geometry the published map was recorded on.
|
| 52 |
|
|
@@ -121,9 +121,9 @@ The map here carries CARLA's own: every signal and stop sign of the 0.10 town, r
|
|
| 121 |
|
| 122 |
- the stop line is the end edge of the approach lanelet, through its two existing nodes -- the nodes every junction lanelet leaving it starts from;
|
| 123 |
- a `traffic_light` regulatory element (a `traffic_sign` one for a stop sign) has that stop line as its `ref_line`, and every junction lanelet (`turn_direction`) leaving the approach references it;
|
| 124 |
-
- each signal head is a `traffic_light` linestring along the bottom of its housing, left to right as the driver sees it, with its `height`, and `light_bulbs` red, yellow and green from the top; a stop sign is a `traffic_sign` linestring of subtype `stop_sign`.
|
| 125 |
|
| 126 |
-
That is 30 traffic-light regulatory elements for CARLA's 15 signals, one per lane they govern, and 3 for its stop signs, on 62 junction lanelets. Its two yield signs have no stop line and are not written.
|
| 127 |
|
| 128 |
</details>
|
| 129 |
|
|
@@ -147,7 +147,7 @@ So a lane change is legal exactly where CARLA paints a broken line -- between sa
|
|
| 147 |
<details>
|
| 148 |
<summary>Provenance</summary>
|
| 149 |
|
| 150 |
-
`lanelet2_map.osm` started as a verbatim copy of [`maps/vector_maps/lanelet2/Town10HD.osm`](https://bitbucket.org/carla-simulator/autoware-contents/src/master/maps/vector_maps/lanelet2/Town10HD.osm) in [carla-simulator/autoware-contents](https://bitbucket.org/carla-simulator/autoware-contents), with SHA-256 `cdd0a28664bcf9b5bd8867584a6fa6a02fa0f75a7c3dbb901ca216128ef86a17`. Three things have been added to it, and nothing of the original geometry changed: the `turn_direction` tags of its junction lanelets, read from CARLA's OpenDRIVE; the lane markings above, written by [`tag_lane_boundaries.py`](https://gist.github.com/hakuturu583/f11582b8ca65a5e2efae9fe884e1dfc0) from a CARLA 0.10.0 server; and the traffic lights, stop signs and stop lines above, written by [`add_regulatory_elements.py`](https://gist.github.com/hakuturu583/f11582b8ca65a5e2efae9fe884e1dfc0) from a CARLA 0.10.0 server. `map_projector_info.yaml` states the projector that file is drawn against.
|
| 151 |
|
| 152 |
`pointcloud_map.pcd` was recorded from a running CARLA 0.10.0 server. A LiDAR with no vehicle attached is placed at every lane sample of the world, 8 m apart and 2.0 m above the road, with an 80 m range. The returns are accumulated in the Autoware map frame and voxel-downsampled at 0.2 m. The recorder is `docker/devcontainer/carla/build-pointcloud-map.py` in [autowarefoundation/autoware](https://github.com/autowarefoundation/autoware).
|
| 153 |
|
|
|
|
| 46 |
| `lanelet2_map.osm` | `pointcloud_map.pcd` |
|
| 47 |
| --- | --- |
|
| 48 |
| [](https://hakuturu583.github.io/simple_lanelet2/embed.html?map=https%3A%2F%2Fhuggingface.co%2Fdatasets%2FAutowareFoundation%2Fcarla-ue5-maps%2Fresolve%2Fmain%2Fautoware_maps%2FTown10HD_Opt%2Flanelet2_map.osm&theme=dark) |  |
|
| 49 |
+
| 168 lanelets, 401 linestrings, 9,426 points, and 33 regulatory elements (15 signals and 3 stop signs). Drawn by [simple_lanelet2](https://github.com/hakuturu583/simple_lanelet2), which also renders it in the browser — **click it to open the file in this dataset in that viewer**. | 1,851,689 points over 368 x 357 x 31 m, voxel-downsampled at 0.2 m. Seen from above, coloured by height — the road surface is the dark ground, and what stands beside it is the buildings, the street trees along the boulevards and the roundabout at the top right. |
|
| 50 |
|
| 51 |
The two are the same town in the two forms Autoware reads it in, drawn to the same frame. What the point cloud shows and the lanelet2 map does not is the reason the point cloud is here at all: this is Unreal Engine 5 geometry, not the 0.9 geometry the published map was recorded on.
|
| 52 |
|
|
|
|
| 121 |
|
| 122 |
- the stop line is the end edge of the approach lanelet, through its two existing nodes -- the nodes every junction lanelet leaving it starts from;
|
| 123 |
- a `traffic_light` regulatory element (a `traffic_sign` one for a stop sign) has that stop line as its `ref_line`, and every junction lanelet (`turn_direction`) leaving the approach references it;
|
| 124 |
+
- each signal head is a `traffic_light` linestring along the bottom of its housing, left to right as the driver sees it, with its `height`, and `light_bulbs` red, yellow and green from the top; a stop sign is a `traffic_sign` linestring of subtype `stop_sign`. A signal is drawn once, and the regulatory elements of all the lanes it governs refer to the same heads, as Autoware's maps share a head between the regulatory elements that refer to it.
|
| 125 |
|
| 126 |
+
That is 30 traffic-light regulatory elements for CARLA's 15 signals, one per lane they govern, sharing their signal's 43 heads, and 3 for its stop signs, on 62 junction lanelets. Its two yield signs have no stop line and are not written.
|
| 127 |
|
| 128 |
</details>
|
| 129 |
|
|
|
|
| 147 |
<details>
|
| 148 |
<summary>Provenance</summary>
|
| 149 |
|
| 150 |
+
`lanelet2_map.osm` started as a verbatim copy of [`maps/vector_maps/lanelet2/Town10HD.osm`](https://bitbucket.org/carla-simulator/autoware-contents/src/master/maps/vector_maps/lanelet2/Town10HD.osm) in [carla-simulator/autoware-contents](https://bitbucket.org/carla-simulator/autoware-contents), with SHA-256 `cdd0a28664bcf9b5bd8867584a6fa6a02fa0f75a7c3dbb901ca216128ef86a17`. Three things have been added to it, and nothing of the original geometry changed: the `turn_direction` tags of its junction lanelets, read from CARLA's OpenDRIVE; the lane markings above, written by [`tag_lane_boundaries.py`](https://gist.github.com/hakuturu583/f11582b8ca65a5e2efae9fe884e1dfc0) from a CARLA 0.10.0 server; and the traffic lights, stop signs and stop lines above, written by [`add_regulatory_elements.py`](https://gist.github.com/hakuturu583/f11582b8ca65a5e2efae9fe884e1dfc0) from a CARLA 0.10.0 server. That script first drew each signal's heads once per lane it governs, two copies on the same points; [`share_signal_heads.py`](https://gist.github.com/hakuturu583/f11582b8ca65a5e2efae9fe884e1dfc0) merged the copies into the ones the regulatory elements now share, and the script itself now draws them once. One thing of the original geometry was changed after all: ten lane bounds stepped from one node to another on the very same spot, a segment of length zero, and Autoware -- which resamples every bound when it loads a map -- made those lanelets' centerlines NaN, which crashed its planner on any route near them. [`merge_duplicate_points.py`](https://gist.github.com/hakuturu583/f11582b8ca65a5e2efae9fe884e1dfc0) merged each such pair of nodes into one; no lanelet moved, and the routing graph is the same. `map_projector_info.yaml` states the projector that file is drawn against.
|
| 151 |
|
| 152 |
`pointcloud_map.pcd` was recorded from a running CARLA 0.10.0 server. A LiDAR with no vehicle attached is placed at every lane sample of the world, 8 m apart and 2.0 m above the road, with an 80 m range. The returns are accumulated in the Autoware map frame and voxel-downsampled at 0.2 m. The recorder is `docker/devcontainer/carla/build-pointcloud-map.py` in [autowarefoundation/autoware](https://github.com/autowarefoundation/autoware).
|
| 153 |
|
autoware_maps/Town10HD_Opt/lanelet2_map.osm
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1881719399090fff1ce9057e8bb4cd0991861e102acd89e4d105747084fe6939
|
| 3 |
+
size 2354525
|