MIT taught a model airplane to avoid collisions with obstacles in cramped spaces

The specialists of the Massachusetts Institute of Technology have created an autonomous model of the aircraft, which can aggressively maneuver in cramped conditions at speeds up to 10 m / s. The aircraft determines its position in space using gyroscopes, accelerometers and a laser rangefinder. The on-board computer runs on a 1.6 GHz Intel Atom processor. The key element of the system is the Gaussian particle filter algorithm optimized for real-time flight.



Modern flying robots are physically capable of making very fast and complex maneuvers. However, until now they could do this only with the help of external positioning and control systems, such as those located at different points in the premises of the video cameras and a computer analyzing their signal. Fully autonomous flight in a confined space while it was possible only for quadros, very light in control and able to move slowly.

Lidar-based navigation systems and particle filters are widely used in robotics, in particular, they are the basis of the famous Google unmanned vehicles. The development of such a system for a small aircraft model is a very nontrivial task, since it is necessary to ensure speed and accuracy for maneuvering in three-dimensional space at a rather high speed, because an airplane, unlike a car, cannot arbitrarily slow down and stop in flight. In this case, the onboard computer should be very easy and economical.
')
PDF with details.