MIT Drones Navigate Obstacles Without Navigation

MIT researchers have developed the SANDO algorithm, enabling drones to autonomously navigate around moving obstacles in unfamiliar 3D spaces without prior map construction.

Unlike traditional systems that attempt to predict the trajectory of every object, the new approach assesses the maximum speed of obstacles and creates a dynamic safety corridor.

The system continuously recalculates the route in real time, ensuring flight safety even when conditions change. Planning, perception, and localization are performed onboard without relying on remote servers.

Testing demonstrated high method efficiency: in simulations, the drone reached its goal without violating constraints, and in 12 actual flights, it successfully avoided all moving objects.

The algorithm is designed for environments lacking maps or where existing ones quickly become outdated, making it promising for rescue operations in rubble and underground mine exploration.

Mathematical safety guarantees hold provided that obstacle speeds and position estimation errors do not exceed specified limits.