CARLA-Air: Unified Simulation for Air-Ground Embodied Intelligence
Researchers have unveiled CARLA-Air, an open-source simulation framework that combines high-fidelity urban driving with physics-accurate multirotor flight in a single Unreal Engine process. This platform resolves the issue of domain separation present in current simulators by integrating both aerial and ground agents into a cohesive physical environment, thus removing the synchronization challenges associated with bridge-based co-simulation. CARLA-Air maintains the original Python APIs and ROS 2 interfaces from CARLA and AirSim, allowing for code reuse without modifications. It offers photorealistic settings featuring traffic that adheres to regulations and multirotor dynamics, all within a unified physics tick and rendering pipeline. This infrastructure facilitates the integration of low-altitude economies, embodied intelligence, and air-ground cooperative systems.
Key facts
- CARLA-Air unifies CARLA and AirSim in a single Unreal Engine process.
- It supports both aerial and ground agents with strict spatial-temporal consistency.
- The platform preserves native Python APIs and ROS 2 interfaces from both simulators.
- It eliminates synchronization overhead from bridge-based co-simulation.
- CARLA-Air delivers photorealistic environments with rule-compliant traffic.
- It addresses domain segregation in existing open-source simulators.
- The infrastructure supports low-altitude economies and embodied intelligence.
- The paper is available on arXiv under ID 2603.28032.
Entities
Institutions
- arXiv