An OpenAI Gym environment wrapper for the CARLA autonomous driving simulator, enabling reinforcement learning research.
gym-carla is an OpenAI Gym environment wrapper for the CARLA autonomous driving simulator. It provides a standardized interface for training reinforcement learning agents to control vehicles in realistic urban simulations. The wrapper handles observation processing, reward calculation, and episode termination specific to driving tasks.
Researchers and developers working on reinforcement learning for autonomous vehicles, particularly those using the CARLA simulator and OpenAI Gym ecosystem.
It simplifies the integration between CARLA's high-fidelity simulation and popular RL frameworks by providing a consistent Gym API, customizable reward/termination functions, and multi-modal observations out of the box.
An OpenAI gym wrapper for CARLA simulator
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Implements the OpenAI Gym interface, ensuring compatibility with popular RL libraries like Stable Baselines and RLlib, as per the project philosophy for reproducible research.
Provides dictionary observations with front camera, LiDAR point clouds, and semantic bird's-eye views, enabling sensor fusion for driving agents, illustrated in README figures.
Allows users to modify reward functions and termination conditions by editing carla_env.py files, supporting tailored RL experiments as described in the usage section.
Seamlessly connects to CARLA's physics-based simulator for high-fidelity vehicle dynamics and urban environments, crucial for accurate autonomous driving research.
Only officially supports Ubuntu 16.04 and is hardcoded for CARLA 0.9.6, restricting use on modern systems and with simulator updates, as noted in installation requirements.
Requires manual steps like conda environment setup, CARLA installation, PYTHONPATH configuration, and separate server launching, making onboarding error-prone and time-consuming.
Primarily designed for research, lacking detailed guides for production use, performance tuning, or troubleshooting beyond basic examples in the README.