explaingit

microsoft/airsim

★ 18,173C++Audience · researcherComplexity · 4/5MaintainedSetup · hard

TLDR

Open-source simulator for drones and autonomous vehicles built on Unreal Engine, enabling AI researchers to test flight control, computer vision, and reinforcement learning without physical hardware.

Mindmap

mindmap
  root((AirSim))
    What it does
      Drone and vehicle sim
      Software-in-loop testing
      Hardware-in-loop testing
      Computer vision mode
    How to use it
      Manual remote control
      Python API
      C++ API
      C# and Java APIs
    Sensor and data
      Camera images
      Depth maps
      Object segmentation
      Sensor readings
    Tech stack
      Unreal Engine
      C++ core
      Python bindings
      Unity experimental
    Use cases
      Train vision models
      Test flight software
      Reinforcement learning
      Autonomous navigation

Things people build with this

USE CASE 1

Train computer vision and machine learning models using synthetic images, depth maps, and segmentation data from the simulator.

USE CASE 2

Test autonomous vehicle flight control software and algorithms without needing expensive physical drones or cars.

USE CASE 3

Collect large datasets of sensor readings and camera feeds for reinforcement learning experiments in a controlled 3D environment.

USE CASE 4

Validate hardware-in-the-loop setups by connecting real flight controller boards to the simulated world.

Tech stack

C++Unreal EnginePythonC#Java

Getting it running

Difficulty · hard Time to first run · 1day+

Requires Unreal Engine installation and compilation from source, plus GPU for realistic simulation.

License could not be detected automatically. Check the repository's LICENSE file before use.

In plain English

AirSim is an open-source simulator for drones, cars, and other autonomous vehicles, built as a plugin for the Unreal Engine game engine (with an experimental Unity version as well). It was originally created by Microsoft Research in 2017 to help AI researchers experiment with deep learning, computer vision, and reinforcement learning in a realistic 3D environment without needing physical hardware. The simulator supports software-in-the-loop testing (running flight control software against the simulator) and hardware-in-the-loop testing (connecting a real flight controller board to the simulated environment). You can control vehicles manually using a remote controller or keyboard, or write code in C++, Python, C#, or Java that communicates with the simulator through a programming interface to collect images, read sensor data, or send movement commands. There is also a dedicated Computer Vision mode where you move cameras through a scene and capture depth maps, surface normal images, or object segmentation data, useful for training machine vision models. Note that this original repository has been archived and is no longer receiving updates. Microsoft has moved development to a new commercial product called Project AirSim, described as an end-to-end platform focused on the needs of the aerospace industry. The archived code remains publicly accessible for reference.

Copy-paste prompts

Prompt 1
How do I set up AirSim in Unreal Engine and write a Python script to control a drone and capture camera images?
Prompt 2
Show me how to use AirSim's Computer Vision mode to generate depth maps and object segmentation data for training a vision model.
Prompt 3
What's the difference between software-in-the-loop and hardware-in-the-loop testing in AirSim, and how do I set up each one?
Prompt 4
I want to train a reinforcement learning agent to navigate a drone in AirSim. What APIs and sensor data should I use?
Prompt 5
How do I connect a real flight controller board to AirSim for hardware-in-the-loop testing?
Open on GitHub → Explain another repo

Generated 2026-05-18 · Model: sonnet-4-6 · Verify against the repo before relying on details.