Technology giant Microsoft has done it once again. Known in the industry for pushing the tech boundaries and inventing the unthinkable, the giant has made news this time for its latest artificial-intelligence (AI) project that gives developers a set of tools to train drones, robots, and several other gadgets to operate autonomously.

Experts comprehend AirSim to be an important part of the AI project, which according to the US-based tech giant offers a very unique and realistic open-source simulator for flying and crashing drones, in order to generate data for training autonomous robots and other vehicles. The simulator for drones is built on Unreal Engine and is available on GitHub for anyone to use and experiment.

What makes AirSim unique is the fact that it is open-source, cross platform and offers support for hardware-in-loop with popular flight controllers such as Pixhawk for physically and visually realistic simulations. It has been developed as an Unreal plugin so that it can be easily dropped into any unreal environment by the user.

According to Microsoft AI & Research, they want AirSim to act as a platform for AI research to experiment with deep learning, computer vision and reinforcement learning algorithms for autonomous vehicles. For this very reason, AirSim also exposes APIs to retrieve data and control vehicles in a platform independent way.

Microsoft states, similar to any other AI project, one of the biggest obstacles that they faced with this particular AI project was to access enough data to be able to successfully train a model for different tasks using a type of deep learning called reinforcement learning or trial and error. In the past, Google's DeepMind has charted success stories by making use of reinforcement learning techniques to train its Artificial Intelligence to master several board games, like Go, as well as numerous 2D Atari games.

[caption id="attachment_113995" align="alignnone" width="700"]Image: Microsoft Image: Microsoft[/caption]

The tech giant is hopeful that its high-fidelity simulated world will be able to provide developers an easy way out to cheaply generate enough data to train an AI system for autonomous flying or driving in the real world. In addition to this, it will also give the developers a chance to crash drones in a safe and secure environment before launching one in the real world.

The simulator in question here is a part of a brand new Microsoft project by the name Aerial Informatics and Robotics Platform. According to Microsoft, the project has been conceived with an aim of bridging the existing gap between simulation and the real world by employing the latest graphics hardware and processors to mimic magnetism, gravity and atmospheric conditions.

Further, the simulator is also completed capable of rendering completely accurate reflections, shadows, and weather-based effects, which can prove to be extremely helpful for a robot that is dependent on computer vision. While currently the platform supports quadrotors, specifically DJI and MavLink-based drones, but according to the tech giant, modules for other vehicles can also be also added on in the near future.

Microsoft is marketing the platform as one that offers its users both time and cost savings.

The platform not only allows the user to create various scenarios, but also enact actions at a rapid rate, for example, hundreds of seconds of real world can be simulated in just one second. It also enables carrying out and studying complex missions that might turn out to be too risky and time consuming in the real world. Further, mistakes and bugs in simulation costs a user virtually nothing. One can crash a vehicle numerous times to get a better understanding of the implemented theories and methods under various conditions without worrying about the cost parameter.

The video below, of AirSim Simulator, shows a simulation of the quadrotor flight in a sub-urban environment. The visuals depict how the camera mounted on the quadrotor perceives the world.

According to Microsoft researcher Ashish Kapoor, the company's aspirational goal is to build systems that can operate in the real world. He strongly believes that, that is the next leap in AI, really thinking about real-world systems.

He also believes that Microsoft's this particular AI project will provide Artificial Intelligence developers all around the world an easier and more practical way of training robots rather than making use of board games with well-defined rules.


Post a Comment

Previous Post Next Post
Like this content? Sign up for our daily newsletter to get latest updates.