Jedi Knights are taught over time to use the Force to augment their skills in battle. Regular old people have to rely on training their reflexes, intuition, and agility to give themselves an edge in the martial arts. But for drones, close quarters combat (com-bot?) is all about real-time kinodynamic planning (essentially real-time obstacle avoidance), which is on display perfectly in the above clip.
The video was made by Ross Allen, a PhD candidate in Aeronautics & Astronautics at Stanford University, whose “current work is focused on training robots on how to avoid obstacles, even at high speeds, while achieving their objectives.” And for his “final test” of the research and technology he’s worked on, he decided to challenge one of his drones to a fencing match. Although it’s not exactly a fair fight as the little buzzing bot—named AgileQuad—doesn’t even have its own blade.
Despite his lack of offensive capabilities, AgileQuad demonstrates its defensive ability to avoid obstacles in real-time at high speeds admirably. With each thrust of Allen’s blade, AgileQuad uses an external motion-capturing system, and super-quick sub-one-second online computations, in order to deduce where it and the blade are in space, and where it needs to go in order to not collide with the blade (or anything else).
While this kind of tech isn’t unprecedented, Allen notes in his research paper — co-authored by Marco Pavone and entitled “A Real-Time Framework for Kinodynamic Planning with Application to Quadrotor Obstacle Avoidance” — that, “This is arguably one of the first, if not the first demonstration of truly real-time kinodynamic planning on a quadrotor system navigating an obstructed environment.” So even though we’ve seen drones effectively avoid obstacles before, Allen’s saying this is the first time that it’s being done in true real-time on a drone.
This kind of obstacle avoidance capability is, of course, not only useful for fencing, nor only useful for drones. In fact, this kind of tech will likely show up on tons of different vehicles and machines, including self-driving cars, planes, spacecraft, and robotic arms, where it will help those things to avoid obstacles and better understand their positions in space. It could also one day be used for battle droids. Real-time kinodynamic planning meets Atlas the robot, anyone?
What do you think about this fencing drone? Are you excited to see these kinds of systems show up on machines more and more in the near future or… battle droids? Let us know in the comments section below!
HT: Popular Science
Images: Ross Allen