AirSim, developed by Microsoft, is an open-source, cross-platform simulator for drones and ground vehicles developed on the Unreal Engine 4 by Epic Games. Several free to use environments have been provided for Airsim, out of which the Neighbourhood and the City are some of the most famous ones. You may have a look at all the currently available environments here. The majority of these environments are suitable to carry out research on Planning, controls, and navigation. There is a lack of appropriate environment which can be extensively used for carrying out research on Simultaneous Localization and Mapping(SLAM) as well as Structure from Motion(SFM).
Here, we make available the Castle Ruins environment, which can be used to carry out research on SLAM, SFM as well as exploration-mapping. The environment depicts an ancient ruins site that is feature-rich and contains multiple locations for place recognition, which can ensure effective loop closures. In addition to this, we provide a ready to use bag-of-words vocabulary specially built using a dataset from this environment which can be directly used in algorithms such as ORB SLAM. Have a look at the overview video shot by simulating a DJI Tello along with PX4 SITL.
Overview of the Castle Ruins SLAM Environment
Here are some of the key features of the environment-
- Spread across an area of 300m x 300m
- Feature-rich and multiple locations for easy place recognition
- Useful for SLAM, SFM, exploration and mapping
- Multi-directional lighting to ensure that the object of interest is well lit
- Support for 2 quadrotors- DJI Tello & AR Drone
If you are new to AirSim, I would suggest going through the tutorials available here. You will have to change the pawn path in the Settings.json file to switch between the drone models. Set the Pawnpath for the default quadrotor to the following-
- For DJI Tello:
- For AR Drone:
This environment was developed during my remote internship under Dr. Aamir Ahmad at the Robot Perception Group, Max Planck Institute of Intelligent Systems. All our work on collaborative SLAM was tested and evaluated in this environment. We will soon be releasing the Castle Ruins Dataset- A monocular cam Dataset for multiple agents that has been recorded in this environment. I will also be uploading a project brief on my work containing the built map of this environment.
Note: The contents for developing this environment have been taken from the freely available resources from the game- Infinity Blade. You can read more about it here.