[mitsudome-r]

Description

Update repository versions in repositories/autoware.repos for the Autoware 1.10.0 release.

Repository	Before	After
core/autoware_core	1.9.0	1.10.0
universe/autoware_universe	0.52.1	0.53.1
launcher/autoware_launch	0.52.0	0.53.0

How was this PR tested?

• vcs import src < repositories/autoware.repos checks out every repository at the new versions
• colcon build passes
• Humble:
  • Planning simulator scenario runs with the updated stack
  • Rosbag Simulation runs with the updated stack
  • AWSIM/CARLA Simulation runs with the updated stack
• Jazzy:
  • Planning simulator scenario runs with the updated stack
  • Rosbag Simulation runs with the updated stack
  • AWSIM/CARLA Simulation runs with the updated stack

[github-actions]

Thank you for contributing to the Autoware project!

🚧 If your pull request is in progress, switch it to draft mode.

Please ensure:

• You've checked our contribution guidelines.
• Your PR follows our pull request guidelines.
• All required CI checks pass before marking the PR ready for review.

[mitsudome-r]

autowarefoundation/autoware_universe#13518 was needed to run rosbag simulation

[mitsudome-r]

Humble Checks

PSim

v1.10.0_psim-2026-10-20_23.09.03-compressed.mp4

Rosbag Replay

v1.10.0_rosbag_sim-2026-10-20_23.14.43-compressed.mp4

AWSIM:
I still saw MRM being triggered time to time, but we have been seeing the behavior since the previous release so ignoring the issue for now. I could due the further investigation and make the patch release once I solve it.

CARLA
Tested with the following:

docker run --rm -it --name carla \
  --runtime=nvidia --gpus all \
  --net=host \
  carlasim/carla:0.9.16 \
  bash ./CarlaUE4.sh -RenderOffScreen -nosound -quality-level=Low

source ~/autoware_release/install/setup.bash
ros2 launch autoware_launch e2e_simulator.launch.xml \
    map_path:=$HOME/autoware_data/maps/Town01 \
    vehicle_model:=sample_vehicle \
    sensor_model:=carla_sensor_kit \
    simulator_type:=carla \
    use_light_weight_sensor_mapping:=True

v1.10.0_carla-2026-10-21_02.19.44-compressed.mp4

[xmfcx]

I tested the 1.10.0 pins from this PR on Jazzy. The planning simulator, rosbag replay and AWSIM demos work. I skipped CARLA for this release.

Planning simulator

Run notes

• Workspace: source build of this PR at 8da95dfeb (autoware_core 1.10.0, autoware_universe 0.53.1, autoware_launch 0.53.0). colcon build finished 502 of 502 packages.
• Platform: Ubuntu 24.04.5, ROS 2 Jazzy, RMW_IMPLEMENTATION=rmw_cyclonedds_cpp, RTX 3090 with driver 595.99.02.
• Command: ros2 launch autoware_launch planning_simulator.launch.xml map_path:=$HOME/autoware_map/sample-map-planning vehicle_model:=sample_vehicle sensor_model:=sample_sensor_kit
• Route: lanelet 143 to lanelet 136, 388 m with two left turns and no traffic lights.
• Result: the route state changed to ARRIVED after 108 s and 387.3 m. The top speed was 4.30 m/s.
• The screenshot is from 54 s after engage, at 15 km/h with 193 m left.
• A first route went through virtual traffic light 10001 on lanelet 44. The vehicle stopped at its stop line and waited for a V2I answer, as designed. The planning simulator has no V2I source.
• Not run here: Humble.

These run notes are AI generated.

Rosbag replay simulation

Run notes

• Same workspace and platform as the planning simulator run, with TensorRT 10.13.3.9-1+cuda13.0 and CUDA 13.0.
• Models: the 20 entries of the artifacts role at 8da95dfeb, downloaded with hf download into a separate directory. Each directory matched its pinned Hugging Face revision.
• Command: ros2 launch autoware_launch logging_simulator.launch.xml map_path:=$HOME/autoware_data/maps/sample-map-rosbag vehicle_model:=sample_vehicle sensor_model:=sample_sensor_kit data_path:=$HOME/autoware_data/ml_models_1.10.0
• RViz: step 3 of the guide, Target Frame base_link, applied through a copy of autoware.rviz with a closer zoom.
• TensorRT built the lidar_centerpoint engines in 7 s and 36 s, and the traffic light engines in 84 s to 365 s. No process died.
• Playback: ros2 bag play ~/autoware_data/recordings/bags/sample-rosbag/ -r 0.2 -s sqlite3 exited with code 0.
• Localization initialized from GNSS, and the NDT alignment succeeded.
• During the 175 s capture: 299 messages on /sensing/lidar/concatenated/pointcloud, up to 24 objects on /perception/object_recognition/objects, and up to 13.1 m/s on /localization/kinematic_state.
• The screenshot is from 101 s into the capture, with 20 objects and the ego vehicle at 22 km/h.
• Not tested: camera-based traffic light recognition, because the sample bag has no images. Humble was not run.

These run notes are AI generated.

AWSIM

Run notes

• Same workspace, platform and models as the rosbag replay run.
• Simulator: AWSIM-Demo-Lightweight v2.0.1, the latest AWSIM release. AWSIM started first, as the quick start says. Its window shows AWSIM v 2.0.1.
• Map: Shinjuku-Map.zip from the AWSIM v2.0.0 release. Its SHA-256 matched the digest of the release asset.
• AWSIM published every topic in the list of the quick start, with the top lidar at 9.9 Hz.
• Command: ros2 launch autoware_launch e2e_simulator.launch.xml vehicle_model:=sample_vehicle sensor_model:=awsim_sensor_kit map_path:=<map dir> data_path:=$HOME/autoware_data/ml_models_1.10.0
• Localization initialized from GNSS. Traffic light recognition published at 8.3 Hz.
• Route: 461 m with two straight crossings, two left turns and three traffic-light lanelets.
• Result: the route state changed to ARRIVED after 189 s and 452.9 m. The top speed was 4.38 m/s. The vehicle stopped three times on the way, for about 21 s, 16 s and 6 s. I did not record the cause of each stop.
• The screenshot is from 163 s after engage, at 15 km/h. The car ahead on the right shows as a detected object with a predicted path.
• Not run here: the HDRP build AWSIM-Demo, and Humble.

These run notes are AI generated.