Getting Started Tutorial


This tutorial refers to the latest versions of CARLA (at least 0.9.5)

Welcome to the ScenarioRunner for CARLA! This tutorial provides the basic steps for getting started using the ScenarioRunner for CARLA.

Download the latest release from our GitHub page and extract all the contents of the package in a folder of your choice.

The release package contains the following

  • The ScenarioRunner for CARLA
  • A few example scenarios written in Python.

Installing prerequisites

The current version is designed to be used with Ubuntu 16.04, Python 2.7 or Python 3.5. Depending on your Python version, execute:

#Python 2.x
sudo apt remove python-networkx #if installed, remove old version of networkx
pip2 install --user -r requirements.txt
#Python 3.x
sudo apt remove python3-networkx #if installed, remove old version of networkx
pip3 install --user -r requirements.txt

Note: py-trees newer than v0.8 is NOT supported.

Running the follow vehicle example

First of all, you need to get latest master branch from CARLA. Then you have to include CARLA Python API to the Python path:

export CARLA_ROOT=/path/to/your/carla/installation
export PYTHONPATH=$PYTHONPATH:${CARLA_ROOT}/PythonAPI/carla/dist/carla-<VERSION>.egg:${CARLA_ROOT}/PythonAPI/carla/agents:${CARLA_ROOT}/PythonAPI/carla

NOTE: ${CARLA_ROOT} needs to be replaced with your CARLA installation directory, and needs to be replaced with the correct string. If you build CARLA from source, the egg files maybe located in: ${CARLA_ROOT}/PythonAPI/dist/ instead of ${CARLA_ROOT}/PythonAPI.

Now, you can start the CARLA server from ${CARLA_ROOT}


Start the example scenario (follow a leading vehicle) in an extra terminal:

python --scenario FollowLeadingVehicle_1 --reloadWorld

If you require help or want to explore other command line parameters, start the scenario runner as follows:

python --help

To control the ego vehicle within the scenario, open another terminal and run:


Note: If you do not wish to automatically (re-)load the CARLA world, you can skip the command line option --reloadWorld

Running all scenarios of one scenario class

Similar to the previous example, it is also possible to execute a sequence of scenarios, that belong to the same class, e.g. the "FollowLeadingVehicle" class.

The only difference is, that you start the scenario_runner as follows:

python --scenario group:FollowLeadingVehicle

Running other scenarios

A list of supported scenarios is provided in List of Supported Scenarios. Please note that different scenarios may take place in different CARLA towns. This has to be respected when launching the CARLA server.

Running scenarios using the OpenSCENARIO format

To run a scenario, which is based on the OpenSCENARIO format, please run the ScenarioRunner as follows:

python --openscenario <path/to/xosc-file>

Please note that the OpenSCENARIO support and the OpenSCENARIO format itself are still work in progress. More information you can find in OpenSCENARIO support

Running scenarios using the OpenSCENARIO format with Global ParameterDeclaration overwrite

python --openscenario <path/to/xosc-file> --openscenarioparams 'param1: value1, param2: value2'

Running route-based scenario (similar to the CARLA AD Challenge)

To run a route-based scenario, please run the ScenarioRunner as follows:

python --route <path/to/route-file> <path/to/scenario_sample_file> [route id] --agent <path/to/agent_file>


python /scenario_runner/srunner/routes_debug.xml /scenario_runner/srunner/data/all_towns_traffic_scenarios1_3_4.json 0 --agent srunner/autoagents/

If no route id is provided, all routes within the given file will be executed.

By doing so, ScenarioRunner will match the scenarios to the route, and they'll activate when the ego vehicle is nearby. However, routes need an autonomous agent to control the ego vehicle. Several examples are provided in srunner/autoagents/. For more information about agents, please have a look into the agent documentation