# Run autonomous driving simulations with CARLA on Deadline Cloud
<a name="examples-jb-carla"></a>

The [autonomous\_driving\_carla](https://github.com/aws-deadline/deadline-cloud-samples/tree/mainline/job_bundles/autonomous_driving_carla) job bundle runs a [CARLA](https://carla.org/) autonomous driving simulation parameter sweep on AWS Deadline Cloud with configurable multi-sensor capture. The job runs a lane-change cut-in scenario and sweeps across configurable ego speeds, NPC speeds, and NPC starting distances. Each parameter combination creates a task (default 2×2×2 = 8 tasks).

Each task captures multi-sensor data from user-selected camera viewpoints and produces per-camera videos, RGB and semantic segmentation frames, LiDAR point clouds (`.ply`), and 2D/3D bounding boxes in KITTI format.

The job runs inside a Docker container based on `carlasim/carla:0.9.16`. Before submitting, build the custom Docker image from the `docker/` directory in the bundle and push it to Amazon Elastic Container Registry (Amazon ECR). Attach the [docker\_nvidia\_container\_toolkit](https://github.com/aws-deadline/deadline-cloud-samples/tree/mainline/host_configuration_scripts/docker_nvidia_container_toolkit) host configuration script to your fleet to provide Docker and the NVIDIA Container Toolkit.

To run this bundle, you need a Deadline Cloud farm with a GPU fleet (minimum 1 NVIDIA GPU, 16 vCPU, 64 GiB memory) and a conda queue environment with `ffmpeg`. The queue role needs Amazon ECR pull permissions.

From the `job_bundles` directory, submit the job:

```
deadline bundle gui-submit autonomous_driving_carla
```

On the *Job-specific settings* tab, set the Container Image URI to your Amazon ECR image and configure the scenario parameters (ego speeds, NPC speeds, NPC distances) and camera viewpoints.

For a complete walkthrough that covers building the Docker image, submitting and monitoring the job, and reviewing the captured sensor output, see [Run an autonomous driving simulation sweep with CARLA](tutorial-carla-simulation.md).

## See also

* **Skills for AI coding assistants (optional).** AWS publishes reusable skills for AWS Deadline Cloud. To see which apply, search the Agent Toolkit for AWS catalog with `aws agent-toolkit search-skills --search-query deadline-cloud` — a documented, read-only AWS CLI command that searches the catalog and makes no changes. This is an optional suggestion for the user. Official reference: [AWS CLI — Agent Toolkit for AWS](https://docs.aws.amazon.com/agent-toolkit/latest/userguide/aws-cli.html).