Cosys-Lab/Cosys-AirSim

What it solves

Cosys-AirSim provides a high-fidelity, physically and visually realistic simulation environment for autonomous vehicles, including drones and cars. It addresses the need for safe, scalable, and detailed testing grounds for autonomous systems without the risks and costs associated with real-world hardware deployment.

How it works

Built as a plugin for Unreal Engine, the simulator integrates directly into Unreal environments. It supports hardware-in-the-loop simulation with popular flight controllers like PX4 and provides extensive API support for controlling vehicles and extracting data. The project expands upon the original Microsoft AirSim by adding advanced sensor simulations (GPU-accelerated LiDAR, Echo/sonar/radar) and enhanced ground-truth data generation for training AI models.

Who it’s for

Researchers and developers working on autonomous vehicle control, SLAM, and robotics, as well as those needing high-quality synthetic data for training computer vision models.

Highlights

• Advanced Sensor Suite: Includes GPU-accelerated LiDAR for higher point density and realistic intensity, as well as Echo sensors for sonar and radar simulation.
• Enhanced Data Labeling: Supports multi-layer annotation and instance segmentation for ground-truth label generation.
• Broad Vehicle Support: Includes skid steering models (e.g., ClearPath Husky) and a specialized BoxCar model for indoor spaces.
• Integration Capabilities: Offers a Matlab API client toolbox and updated ROS/ROS2 implementations for seamless robotics framework integration.
• Realistic Visuals: Features camera distortion effects like chromatic aberration, motion blur, and lens distortion.