Navigation auf uzh.ch
In a recent study published in Science Robotics, Davide Scaramuzza and his team trained an autonomous quadcopter to fly at up to 40 kilometers per hour through unknown environments without colliding with obstacles. The drone's neural network learns to avoid obstacles by watching a “simulated teacher”, i.e., an algorithm that flew a simulated drone in a virtual environment.
This simulated teacher cannot be used outside of the simulation, but its data are used to teach the neural network how to find the optimal trajectory based on the quadcopter's sensors.
The next step will be to improve the system and develop faster sensors that provide more environmental information in less time, so that the drone can fly safely even at speeds above 40 kilometers per hour.