FlyChams: plataforma open-source escalable para simulación aérea fotorrealista
TL;DR
This work introduces FlyChams, a high-performance, self-hosted, open-source simulation platform, built on Unreal Engine 5 and a custom fork of AirSim, which accommodates demanding scenarios ---including multiple agents, numerous simultaneous video streams, and high resolutions thanks to a hardware-accelerated streaming layer.
Abstract
Photorealistic simulation is essential for aerial robotics research, particularly for perception tasks involving multi-vehicle systems. However, the current open-source simulation paradigm falls short; benchmark platforms have either been discontinued or transitioned to closed-source models, such as AirSim ---now AirGen---. Meanwhile, newer alternatives based on Isaac Sim remain immature for aerial fleets. This work introduces FlyChams, a high-performance, self-hosted, open-source simulation platform. Built on Unreal Engine 5 and a custom fork of AirSim, it is governed by the PX4 autopilot and integrated with ROS 2. Its design accommodates demanding scenarios ---including multiple agents, numerous simultaneous video streams, and high resolutions--- thanks to a hardware-accelerated streaming layer. Finally, we present a search and rescue mission to demonstrate the platform's capabilities and evaluate its scalability.