The scientific project OSTAR is a set of software tools for automotive simulation. OSTAR enables a simulation where vehicles in CARLA are controlled by external models. Theses models are packaged with FMI and use OSI messages for communication. The integration of theses models to CARLA is given by the GroundTruth, SensorView and TrafficUpdate messages.
This Quickstart shows how to use OSTAR. Further customization is possible and can be found in the documentation. OSTAR is an implementation of the SetLevel generic simulation architecture. OSTAR also supports SUMO to generate traffic scenarios.
System capable to run docker containers and a GPU to support CARLA. Download Git Large File Storage (LFS) for example scenarios with custom maps.
apt install git git-lfsIn the setup.sh script, you can choose to either download or build SCS (Single Container Simulation) or build DS (Distributed Simulation).
It is recommended to start with the download of SCS since the first build of SCS takes up to one hour.
Only move to a distributed simulation if neccessary. The distributed simulation starts OSTAR component in separate docker containers.
This is useful, if models only work in custom environments.
git clone https://github.com/DLR-TS/OSTAR-Quickstart.git
cd OSTAR-Quickstart
chmod +x *.sh
./setup.sh --dockerTo run the simulation choose between the provided examples. You can select the scenarios interactively or as a path by parameter.
./run.sh --visualAvailable runtime parameter are:
[scenario directory] directory with scenario files
[image] optional image selection
[--visual] to open CARLA window
[--verbose] to enable verbose logs
The simulation runs the scenario and will generate OSI trace files in a new output directory. Images are stored in a separate directory.
Check out the OSTAR documentation at OSTAR Quickstart.
[email protected] [email protected] [email protected]
This software was originally developed as part of SetLevel.