AirGen is a next-generation simulator focused on enabling General Robot Intelligence through large-scale data generation and evaluation for machine learning. Built on top of AirSim, AirGen supports various types of robots, including ground and aerial robots, in diverse environments such as urban settings, industrial areas, and natural landscapes.

To increase the ease of development, a Python API is integrated with GRID to allow for seamless development while also being exposed to allow the user full control over the simulation environment and robotic agents from GRID. In this section we will cover the key features of Airgen as well as document the API.

The client described in Robots is fundamental to interacting with the airgen simulation; it is highly recommended to read this page first

​
Key Features

  • Unreal Engine 5 Ready: AirGen is built on top of Unreal Engine 5, enabling you to leverage the latest features in Unreal Engine such as Nanite, Lumen, and more.

  • Geospecific Environment Support: Dive into real-world terrains and photorealistic tilesets with AirGen’s support for geospecific environments powered by the Cesium platform. Experience a seamless blend of the virtual and real world, delivering unparalleled experiences for your simulation needs.

  • Advanced Path / Trajectory Planning: Leveraging features such as occupancy maps, signed distance fields, path planning algorithms, and minimum snap trajectory generation, AirGen enables generating optimal and collision-free trajectories at scale.

  • Large-Scale Data Collection: AirGen’s support for large-scale data collection allows you to collect virtually infinite amounts of photorealistic multimodal data, unlocking the training of generalizable perception-action models.

  • Digital Twin Import: AirGen supports importing meshes or scenes at runtime, enabling you to simulate real-world context within the simulation.

Was this page helpful?