Self-driving research sim moves to Unreal 5, and so must your GPU

What it does

CARLA is an open-source simulator built on Unreal Engine 5.5 for autonomous driving research. It provides open digital assets—urban layouts, buildings, and vehicles—along with flexible sensor suites and environmental conditions so researchers can train and validate driving algorithms in a virtual world. The project includes a Python API and sits at the center of an ecosystem that includes benchmarking tools, ROS bridges, and scenario runners.

The interesting bit

Most simulators ship code and leave you to build the world; CARLA includes purpose-built open digital assets licensed under CC-BY, giving researchers a ready-made virtual city rather than an empty grid. The project is currently straddling two major engine versions, with the UE5.5 branch existing in parallel to the UE4.26 line and carrying significant differences.

Key highlights

• Ships with open digital assets (urban layouts, buildings, vehicles) under a CC-BY license
• Supports flexible specification of sensor suites and environmental conditions
• Provides a Python API and integrates with ROS, AutoWare, and reinforcement-learning pipelines
• Ecosystem includes leaderboards, scenario runners, and imitation-learning benchmarks
• Core simulator code is MIT-licensed

Caveats

• Hardware requirements are steep: the README recommends an RTX 3070 or better, 32 GB+ of RAM, and 16 GB+ of VRAM
• UE5.5 version is locked to Ubuntu 22.04/24.04 or Windows 11 and explicitly will not work on Ubuntu 20.04 or Windows 10
• Building from source requires linking your GitHub account to Epic Games and using your credentials to authorize download of the Unreal Engine 5.5 repository
• Unattended setup is currently unavailable on Windows

Verdict

Autonomous driving researchers and AV stack developers who need a configurable virtual proving ground with detailed environments should start here. If you lack high-end NVIDIA hardware or were hoping for a lightweight, turn-key simulator, look elsewhere.