Drones are getting smarter and smarter, adding tasks like navigating trails and dodging obstacles to their arsenal. With industrial applications like delivering parcels and inspecting powerlines they will need all the smarts they can muster. The latest innovation on the horizon is that quadcopters now come in sword-fighter resistant form…
To demonstrate “agile avoidance maneuvers for high speed obstacles” Ross Allen and Marco Pavone headed leftfield and confronted their drone with a fencer, cutting and thrusting in the quadcopter’s direction – fortunately to no avail, as can be seen in the footage above from Stanford’s Autonomous Systems Laboratory.
As well as dodging the fencer’s blade, the system incorporates a machine-learning element, so these little drones will just be getting smarter and smarter in the future, and will be able to find their way around static and moving obstacles.
The research, carried out by Allen and Pavone of Stanford University’s Department of Aeronautics and Astronautics department, intends to achieve “exploration through machine learning, decision making through optimal control, precomputation when possible.”
Currently, the drone’s responsiveness is based on cameras mounted around the space, but this will evolve to rely on visual, laser and ultrasonic sensors placed on the actual drone in future.
Just to be clear, this drone can’t actually fence – wielding a sabre would be an entirely separate piece of research, and is beyond the remit of Allen and Pavone’s research (which emphatically doesn’t feature sword-wielding, autonomous quadcopters at all). That said if Amazon’s delivery drones can circumnavigate swordsmen, then they might have a much better chance of dealing with the general public, and everything else that gets thrown at them.
If you’d like to read Allen and Pavone’s research paper, you can find it here: “A Real-Time Framework for Kinodynamic Planning with Application to Quadrotor Obstacle Avoidance“.