Microsoft has unveiled a simulator for training autonomous drones, powered by Microsoft Azure to allow for millions of complex flight simulations per second.
Project AirSim is the successor to Microsoft’s earlier AirSim, beginning its life as a research project based in Unreal, but requiring an extensive understanding of machine learning and coding.
The newly-announced iteration promises an end-to-end experience that will allow users an easy but powerful tool to train AI models within a 3D environment.
This environment is drawn from Microsoft’s own Bing Maps as well as other sources of geographic information and can represent specific locations such as New York City or London, as well as custom environments to suit a customer’s individual training needs.
Using Azure to run the simulator allows tests to be run on the cloud at scale, in turn accelerating the training of customer models. The availability of pre-trained AI building blocks also promises to speed up development and prevents vehicle developers from having to expend time on redundant research.
Currently, the project is undergoing development with regard to its weather and physics models. To this end, Microsoft is collaborating with Ansys to implement their high-fidelity physics simulations within Project AirSim.
Software company MathWorks is also contributing to the project, utilising their graphical programming environment Simulink to allow user customisation of the physics environment.
“Everyone talks about AI, but very few companies are capable of building it at scale,” explained the engineering lead for Project AirSim, Balinder Malhi.
“We created Project AirSim with the key capabilities we believe will help democratize and accelerate aerial autonomy – namely, the ability to accurately simulate the real world, capture and process massive amounts of data and encode autonomy without the need for deep expertise in AI.”
Microsoft has indicated that it will work with global civil aviation regulators on deciding Project AirSim’s place in the certification of autonomous systems, floating the idea of a series of scenarios generated within AirSim that could prove a benchmark for autonomous vehicles.
Aerospace manufacturer Bell has already used AirSim to prepare their Autonomous Pod Transport (APT) aircraft for NASA’s Systems Integration and Operationalization (SIO) project, which sought to demonstrate the viability of unmanned aircraft.
The COO's pocket guide to enterprise-wide intelligent automation
Automating more cross-enterprise and expert work for a better value stream for customers
Autonomous aircraft are an increasing item of interest in the public as well as the private sector. The UK government has announced £105.5 million to be allocated to projects such as Open Skies Cornwall, which will work with the Royal Mail and NHS to make deliveries to the Isles of Scilly using drones, as well as Skyway, which will use drones to survey infrastructure to inform construction projects.
With Project AirSim looking to tackle transport, delivery, and aerial infrastructure inspection, the scope for development in the sector looks wide open and full of promise. It is available now as a limited preview, with access expanding in the near future.
