Experimenting With Robotic Movement

kodiaktau writes "Roboticists with EPFL are playing with new methods of locomotion for robots, modeled after grasshoppers, bats and other non-traditional forms of movement, including leaping and gliding."

Comments

Jumpgliding?

[frank_adrian314159]

Looks more like jumpcrashing to me.

Re:Jumpgliding?

[spectrokid]

An object which with relative ease can go down a slope or staircase, but has no chance in hell to get back upstairs? Mmmm, interesting, but might meet stiff competition from this other wonderful robot called "rubber ball"...

Re:Traditional for whom ?

[Missing.Matter]

Fned is correct. Most robots you see today are known as fully actuated systems. Examples of fully actuated sytems are robotic arms or differential drive robots, or even most humanoid robots. The reason for this control of a fully actuated systems is relatively easy and predictable; given a configuration of a fully actuated robot, we can transition the robot to any other state within its state space. This is why most humanoid robots don't look right when they walk; they strive to retain full control authority and balance at every step, whereas human's are really in a constant state of imbalance as we walk.

The problem is all the very interesting systems out there are under actuated, like the walking human. That is, they have more degrees of freedom than ways of exerting control. In nature, things like birds, fish, insects, and even bats and grasshoppers are under actuated. They have extraordinary mobility, but our robotic equivalents fail miserably. I think it's safe to say that most of the exciting problems in motion planning and control in robotics are in the area of underactuated systems.

Robotics is dead for a while

[Spiked_Three]

This post has been up for hours, and there are 17 comments. Robotics is so dead it isn't funny. Reminds me of AI in the 80's - it's taken 30 years for it to make some comeback. Same thing will happen in robotics. Check back in 25 years. The processing power just isn't there at the moment. And yes it takes a ton of processing power to do anything useful with 3D images - like reacting fast enough and in a way that doesn't kill humans. Plenty of toys at the moment following black lines on the floor, but until we can process images quickly, robotics isn't going anywhere. It will be stuck in university research for many years. BTW, researching nature IS probably a good idea - and when a 1/4 lb computer can process as well as a grasshopper we might be making progress. We can't even do it with a 1 ton computer at this time.