MIT's new drone is never sure where it is - and this helps it avoid collisions

Oddly enough, but the control systems of modern drones "suffer" from excessive accuracy. Often hundreds of measurements are required for the UAV to know exactly where it is in space. In case of any inaccuracy, he simply "swoops" on the nearest obstacle. However, loading such a volume of information into a small drone is not always possible.

Developed by the Massachusetts Institute of Technology (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL), the new NanoMap mapping system strikes a harmonious balance between accuracy and speed. With its help, drones can move at speeds up to 32 km / h in places "populated" by many objects, for example, in a dense forest or in a city.

Unlike other mapping systems (for example, SLAM), the NanoMap algorithm is based on the principle of uncertainty in the process of movement. The system measures the depth of the surrounding space, after which it forms a single 3D image from separate "pictures". Now the drone can more accurately assess the environment and predict its flight in relation to the objects encountered.

If without the NanoMap system, every fourth test flight of the drone ended in a collision, then with its installation, the number of collisions decreased to 2%.