docs(docker-new): add compose examples (basic + awsim + planning-sim)

[xmfcx]

• Parent Issue: Simplify the docker images and workflows #6852
• basic/ — three dev-shell compose files picking the right universe-devel-* image for the host GPU story: dev-cpu.compose.yaml (LIBGL_ALWAYS_SOFTWARE=1, no GPU), dev-dri.compose.yaml (/dev/dri passthrough for Mesa iris/radeonsi/nouveau), dev-nvidia.compose.yaml (runtime: nvidia + universe-devel-cuda-jazzy). All three forward $DISPLAY, mount ~/autoware_map / ~/autoware_data, and remap the in-container aw user to the host UID/GID.
• demos/planning-simulator/ — auto-launches planning_simulator.launch.xml with the sample map on universe-jazzy. Base docker-compose.yaml ships software rendering; docker-compose.dri.yaml and docker-compose.nvidia.yaml are overlays merged via -f base -f overlay for Intel/AMD or NVIDIA hardware acceleration.
• demos/awsim/ — bridges to the AWSIM Unity simulator via network_mode: host and launches e2e_simulator.launch.xml on universe-cuda-jazzy. See docker-compose.yaml.
• Entrypoint convention. Demo command: blocks end with exec bash so the container drops to an interactive shell after the launch exits, instead of disappearing.

Why

The docker-new images landed without any user-facing launch examples — contributors had to reverse-engineer which tag to use, how to forward X11, when to add runtime: nvidia vs /dev/dri, and how the HOST_UID/HOST_GID entrypoint remap works. These compose files are the canonical "which knob for which host" reference and pair the three GPU passthrough strategies with the three image flavors (universe-devel, universe, universe-cuda) we actually ship.

---

• 6 of 6 PRs splitting ci(docker-new): split base-cuda layer and restructure CI pipelines #7025. Final PR in the stack.
• Depends on ci(docker-new): split base-cuda layer and restructure CI pipelines #7033 (merged).

Test plan

• Basic dev-shell on an NVIDIA host — expect glxinfo -B to report NVIDIA Corporation:

  xhost +local:docker && HOST_UID=$(id -u) HOST_GID=$(id -g) docker compose -f docker-new/examples/basic/dev-nvidia.compose.yaml run --rm autoware bash -c "glxinfo -B | grep -E 'OpenGL (vendor|renderer)'"

• Basic dev-shell on an Intel/AMD host — expect Mesa Intel... or AMD Radeon..., not llvmpipe:

  xhost +local:docker && HOST_UID=$(id -u) HOST_GID=$(id -g) docker compose -f docker-new/examples/basic/dev-dri.compose.yaml run --rm autoware bash -c "glxinfo -B | grep -E 'OpenGL (vendor|renderer)'"

• Basic dev-shell CPU fallback — expect llvmpipe:

  xhost +local:docker && HOST_UID=$(id -u) HOST_GID=$(id -g) docker compose -f docker-new/examples/basic/dev-cpu.compose.yaml run --rm autoware bash -c "glxinfo -B | grep -E 'OpenGL (vendor|renderer)'"

• Planning simulator launches on base compose (software rendering). With ~/autoware_map/sample-map-planning present, expect rviz to render the vector map:

  xhost +local:docker && HOST_UID=$(id -u) HOST_GID=$(id -g) docker compose -f docker-new/examples/demos/planning-simulator/docker-compose.yaml run --rm planning-simulator

• Planning simulator NVIDIA overlay merges cleanly — expect runtime: nvidia in the resolved config:

  docker compose -f docker-new/examples/demos/planning-simulator/docker-compose.yaml -f docker-new/examples/demos/planning-simulator/docker-compose.nvidia.yaml config | grep -E "runtime|NVIDIA_"

• Planning simulator DRI overlay merges cleanly — expect /dev/dri in devices::

  docker compose -f docker-new/examples/demos/planning-simulator/docker-compose.yaml -f docker-new/examples/demos/planning-simulator/docker-compose.dri.yaml config | grep -E "devices|/dev/dri"

• AWSIM demo config resolves with network_mode: host and runtime: nvidia:

  docker compose -f docker-new/examples/demos/awsim/docker-compose.yaml config | grep -E "network_mode|runtime"

• UID/GID remap works — inside any of the dev shells, expect the aw user's UID/GID to match the host:

  HOST_UID=$(id -u) HOST_GID=$(id -g) docker compose -f docker-new/examples/basic/dev-cpu.compose.yaml run --rm autoware bash -c "id -u && id -g"

[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.