Experiments on synthetic data generation within the CARLA simulator

This project was developed as part of the Cognitive Architectures research line from the Hub for Artificial Intelligence and Cognitive Architectures (H.IAAC) of the State University of Campinas (UNICAMP). See more projects from the group here.

Repository Structure

All experiments conducted are organized under the experiments directory. Each subfolder corresponds to a specific test scenario and includes a dedicated README.md file with detailed information about the setup, methodology, and results.

Experiments

• 1_vehicle_data
  Collection of synthetic sensor data from a simulated autonomous vehicle in a controlled environment. Sensors include IMU, GNSS, and camera modules.

• 2_pedestrian_data
  Preliminary tests with pedestrian agents equipped with sensors. The goal was to evaluate the feasibility of capturing motion data from walking agents and identify limitations in CARLA’s sensor support for pedestrians.

• 3_custom_routes
  Implementation of custom navigation routes using waypoints and intermediate stops. The vehicle is guided by a navigation agent (e.g., BasicAgent) to follow a defined path.

• 4_vehicle_behaviour
  Comparative analysis of different driving behaviors (Cautious vs. Aggressive) using CARLA’s BehaviorAgent. The experiment assesses how driving style affects sensor data and trajectory dynamics.

• 5_custom_maps
  Import and generation of custom maps based on OpenStreetMap (OSM) data. Includes an attempt to enrich OSM maps with elevation data (DEM/MDT) for more realistic terrain modeling.

Auxiliary Components

• agents/
  Contains navigation agents used to control vehicle behavior during simulations.

• modules/
  Provides reusable components for sensor setup, data recording, and post-processing (e.g., plotting and CSV export).

Each experiment is self-contained and can be executed independently. Refer to the corresponding README.md files for specific instructions.

---

Dependencies / Requirements

System Requirements

The following requirements should be fulfilled before installing CARLA:

• Operational System: Windows 10/11 or Linux systems.

• GPU: ≥ 6 GB VRAM (dedicated GPU recommended)

• Disk space: ≥ 20 GB

• Two TCP ports available: By default, CARLA uses ports 2000 and 2001. Ensure these ports are not blocked by firewalls or other applications. If you plan to use the Traffic Manager, also make sure that ports 8000 and 8001 are avaliable.

Requirements

• Python: Python 3.7.

• Pip. version 20.3 or higher.

python3 --version
pip3 --version

python3.7 -m venv .venv
source .venv/bin/activate
pip3 install -r requirements.txt

Installation / Usage of CARLA 0.9.15 (via GitHub Package)

1. Download the package

1. Visit the 0.9.15 release page and download:
   • CARLA_0.9.15.tar.gz
   • AdditionalMaps_0.9.15.tar.gz

2. Extract the package

sudo mkdir -p /opt/carla-simulator
sudo tar -xzf CARLA_0.9.15.tar.gz -C /opt/carla-simulator
cd /opt/carla-simulator

3. Import additional assets (maps)

Place AdditionalMaps_0.9.15.tar.gz into the Carla root and run:

cd /opt/carla-simulator
tar -xzf ../AdditionalMaps_0.9.15.tar.gz
./ImportAssets.sh

4. Install the Python client

pip3 install carla==0.9.15
pip3 install -r PythonAPI/examples/requirements.txt

5. Run the CARLA server

cd /opt/carla-simulator
./CarlaUE4.sh

---

Acknowledgements

This project is part of the Hub for Artificial Intelligence and Cognitive Architectures (H.IAAC- Hub de Inteligência Artificial e Arquiteturas Cognitivas). Project supported by the brazilian Ministry of Science, Technology and Innovations, with resources from Law No. 8,248, of October 23, 1991.