New Virtual Reality System to Train Self-Flying Drones


MIT engineers developed novel Virtual Reality (VR) training system for drones, enabling vehicle to visualize virtual environment while flying in an empty physical space

The system named as Flight Goggles can significantly reduce the number of crashes experienced by drones in actual training sessions. New technology aims to build a virtual testing ground for drones to face various environments and conditions, which can occur during flying. The research will be presented next week in Brisbane, Australia, at the IEEE International Conference on Robotics and Automation. These drones are designed to fly through a large empty real-life room, which is tracked by motion-capture cameras that note orientation of the drone in three-dimensional space. “The drone will be flying in an empty room, but will be ‘hallucinating’ a completely different environment, and will learn in that environment,” said led associate professor Sertac Karaman.

Furthermore, a computer running an image-rendering system transmits interactive first-person-view video of an obstacle-filled environment to the drone, which is visualized by it in place of the feed from its onboard camera. Each virtual obstacle is assigned as physical location in the actual room. Moreover, the aircraft attempts to avoid the computer-generated obstacles, with the motion-capture cameras. Upon failure of doing so, it is considered as drone can crash into anything in the real world.

As a part of examination, Flight Goggles trained a drone to fly through a virtual window about twice as large as the aircraft itself. Navigation algorithm self-tuned itself with trial and error, throughout the course of these training sessions. The drone subsequently utilized that algorithm to fly through an actual window the same size and location as the virtual one, guided by its onboard camera. The drone successfully crossed the window 119 times, only crashing or requiring assistance an additional six times, which was part of the intended learning process.


Comments are closed.