Typical robot motions are linear and tend to move quickly and stop in between separate actions, which tends to look very mechanical. This is a problem when your robot has been designed to work closely with people. Perhaps with the help of traditional animators, a robot programmer could learn a thing or two about more fluid, life-like motions, but even a nice animation or motion routine can become mechanistic if repeated every time a specific action is called for. Assistant Professor Andrea Thomaz and Ph.D. student Michael Gielniak at Georgia Tech are working on more human-like movements using Simon. They found that by mixing different motions routines so that they flow into one another smoothly not only made them more life-like, but that human observers were better able to understand what the robot was doing. Additionally, they plan to create a way to subtly tweak a motion routine so that it isn’t exactly the same each time it is performed, which should add nuance to Simon’s existing repertoire.
Earlier we got a look at some pioneering work being done at Japan’s Advanced Institute of Industrial Science and Technology (AIST) by Sebastien Lengagne, Karim Bouyarmane, and Abderrahmane Kheddar, which allows the full-size humanoid HRP-2 Promet to navigate its environment by steadying itself on objects rather than avoiding them. A new video shows how the HRP-2 can use this skill to lean onto a desk and drop a ball into a wastebasket underneath it. A relatively simple task like this might not seem all that impressive, but without the ability to lean on the desk it would be made much slower. It would have to kneel down in front of the desk, and position itself in such a way that its arm would not collide with the desk when reaching to the basket.