usdotfhwastol/scenario-runner repository overview

⁠CARLA ScenarioRunner scenarios

This directory contains custom ScenarioRunner⁠ scenario configurations to facilitate integration testing. This still a work in progress, so the scenarios serve more as example references.

⁠Getting started

ScenarioRunner's code resides in the srunner Python package, and the scenario_runner.py script is responsible for launching scenarios. If we want to use ScenarioRunner for custom CARMA scenarios, we have to overcome three technical hurdles:

1. the srunner package is unavailable in PyPi⁠;
2. we have to embed our scenario definitions and configurations within srunner; and
3. ScenarioRunner uses on the CARLA Python API⁠ internally, which we have to install separately.

You have three options for getting ScenarioRunner working with the custom CARMA scenarios, each described below.

⁠Manual installation

Check out the official documentation on how to install ScenarioRunner, the CARLA Python client library. Their instructions also show how to add custom scenario definitions (in the scenarios/ directory) and configurations (in the examples/ directory) to the installation.

⁠Convenience script

You can use the convenience script (install_carma_scenario_runner) provided in this directory to download and install a specific ScenarioRunner version to a desired location. The script will also replace the provided scenarios with the ones specific to CARMA.

The example below shows how you can install ScenarioRunner version 0.9.10 to the /opt/scenario_runner directory:

$ ./install_carma_scenario_runner --prefix /opt 0.9.10

⁠Docker (recommended)

You can build a Docker image that packages ScenarioRunner, the CARMA scenarios, and the CARLA Python API. The Dockerfile included in this directory handles all the required installation steps.

After building the image, you can run ScenarioRunner as shown in the below example:

$ docker run --rm -it usdotfhwastol/carma-scenario-runner:0.9.10 \
    --scenario FollowLeadingVehicle_1 --reloadWorld

The container automatically invokes the scenario_runner.py script, so you need only to specify the desired scenario.