Olympic Movement & Robotic Design
The Olympic Winter Games are a showcase of human movement and athletic achievement. Raffaello D'Andrea of the Swiss Federal Institute of Technology draws inspiration from Olympic athletes with the quadrocopter, a flying robotic device that has the ability to learn and improve its performance over time. "Science and Engineering of the 2014 Olympic Winter Games" is produced in partnership with the National Science Foundation.
LIAM McHUGH, reporting: The speed of Ted Ligety.
TED LIGETY (Alpine Skiing Gold Medalist): You have to be pretty dialed in on where you're going.
McHUGH: The power of Julie Chu.
JULIE CHU (Hockey Silver & Bronze Medalist): The physical part of the game is something that we embrace.
McHUGH: The artistry of Meryl Davis and Charlie White.
MERYL DAVIS (Figure Skating Silver Medalist): There's a lot of complexity involved in making each element as efficient as possible.
McHUGH: Every four years the Olympic winter games are a showcase of human movement and athletic achievement, with performances so dazzling they have the ability to amaze and inspire.
RAFFAELLO D’ANDREA (ETH Zurich): Athletes push the boundaries of human abilities, well; we want to push the boundaries in a similar way with our research.
McHUGH: Raffaello D’Andrea is a professor of dynamic systems and control at the Swiss Federal Institute of Technology in Zurich, Switzerland. He was supported by the National Science Foundation while on the faculty at Cornell University in Ithaca, New York.
McHUGH: He and his colleagues at ETH Zurich are at the forefront of robotic engineering, working with machines that mimic human movement, such as the small flying vehicle known as the quadrocopter, or quad. It's similar to a helicopter but with four independent rotors that allow it to move with great agility and acrobatic ability.
D’ANDREA: A lot of times what we end up doing looks very similar to what human beings would do when playing sports.
McHUGH: But what makes the quad so innovative isn't just the way it moves, it's the machine's ability to learn and improve its performance.
D’ANDREA: It's similar to what athletes do when they practice. They do the same task over and over again. And by doing it they can get better over time. Our machines do the same thing.
McHUGH: To showcase the quad's abilities, D’Andrea's team sets it up to complete a series of experiments that are similar to how Olympic athletes move. They begin with a slalom course to teach the quad how to maneuver through a series of poles like skier Ted Ligety.
Man #1: Are you ready?
Man #2: Yep.
Man #1: Ok, let's start it!
McHUGH: Using mathematical models to roughly tell them how the quad should control itself, D’Andrea's team writes algorithms, or a sequence of instructions that help the quad figure out how to execute its task.
D’ANDREA: The process for us when designing a robot is we really think of a task first. And then we figure out, you know, what kind of design does the robot have to have in order to do that?
McHUGH: The test is carried out in a special arena equipped with cameras that are part of the quad's control systems. These systems measure the quad's physical progress, and provide feedback that helps the quad make adjustments during further test runs.
D’ANDREA: And over a period of time, five or six iterations, what it couldn't do at-- in the very first trial, it can do now quite well.
McHUGH: This method of machine learning also applies if the quad wants to intercept and bounce a ball back to where it was thrown, much like a hockey player Julie Chu passing the puck to a teammate.
D’ANDREA: They first have to put the puck in an open space, but knowing exactly where the other player will come by and hit the puck perhaps to another player. Our vehicles are doing the same thing. They are seeing where the ball is flying and then they're executing algorithms that figure out where should I intercept the ball so that when I hit it, it goes exactly where I want it to go.
McHUGH: In ice dancing, Meryl Davis and Charlie White must react in real time to each other, and make quick adjustments to produce a near flawless performance.
D’ANDREA: The quadrocopters are similar in that way. But at the same time, they must react to changes in the conditions if one of their partners doesn't do what they're supposed to be doing.
McHUGH: In this scenario, the researchers create a mathematical model of a quad combined with an object it needs to balance. Then, with feedback from its control systems, algorithms are built so the quad can complete its task.
D’ANDREA: If you find that the machine needs improvement, you're going to redesign it, you do more simulation; develop new algorithms and the process builds up until you have a machine that does incredible feats.
McHUGH: People often marvel at how effortlessly athletes and the quad perform their tasks. Both robot and Olympian have something in common: they need to practice their moves countless times before it looks easy.
D’ANDREA: It's very motivating to watch professional athletes and world class athletes at the top of their game. It certainly motivates us to do similar things with our machines.
McHUGH: Whether it's skiing, hockey, ice dancing or other sports, the athletes of the 2014 Olympic Winter Games will serve as a source of inspiration for D’Andrea and his team. They in turn hope their research will help lay the groundwork for the design of more responsive and agile robotic devices.