Microsoft Makes Its Drone AI Training Simulator Kit Available As Open Source Software

Microsoft has open sourced a beta version of a simulator system its researchers use to train and test software for the autonomous operation of robots, drones and other gadgets.

The suite of tools which form the Redmond company’s Aerial Informatics and Robotics platform, are now available on GitHub.

 Through the use of the Unreal graphics engine, commonly  used in mainstream computer games, the tools enable users to quickly write code to control aerial drones and other autonomous gadgets in a realistic simulated virtual environment in order to safely test its performance and gather data for training such autonomous systems for use in the real-world.

Open-source drone training

Testing autonomous drones in the real-world can be a tricky process as a collision with objects such as a building, bird, other drones could not only damage the drone but also pose a risk to people, animals or objects in the test area.

Using a virtual environment to test the autonomous code and algorithms presents a much safer and cost-effective way of testing autonomous software, as well as enabling testing to be carried out at higher volumes, potentially speeding up the time to market for gadgets with autonomous capabilities.

“The aspirational goal is really to build systems that can operate in the real world,” said Microsoft researcher Ashish Kapoor, noting that the tools have potential to be used to train driverless cars as well as drones and robots in a fashion that addresses their use in real-word situations.

“That’s the next leap in AI, really thinking about real-world systems,” he said.

Using such simulation tools differs from the more traditional process of training artificial intelligence (AI) based systems, which tend to be focussed on training autonomous software to be more effective in artificial environments, such as making the best moves in board games, as seen with Google DeepMind’s AlphaGo AI.

Simulators for training AIs and autonomous systems are not new, but Microsoft boasts that its take is more advanced and the ability to render realistic looking environments makes it a better platform for training systems that are reliant on computer vision algorithms.

“If you really want to do this high-fidelity perception work, you have to render the scene in very realistic detail – you have sun shining in your eyes, water on the street,” said Shital Shah, a principal research software development engineer who was a key developer of the simulator.

Microsoft’s researchers plan to add more tools to the platform, but in the short-term hope that by making the current set of tool available as open source software Microsoft will help push the field of AI and autonomous system development forward and look to establish standardised protocols and regulation surrounding the creation and use of AI agents and systems.

The Redmond company appears to be making more of its software tool available as open source code, such as its machine learning toolkit and its Powershell system administration tool for Windows, Linux and Mac OS X.

Try our open source quiz!