Moving robot with stereoscopic live vision

Yeah, I guess anything is possible with the proper development team.

you cannot move a robot with a gyroscope. a gyro is a sensor that detect acceleration (and by calcul, movement, and by calcul again relative position (from an orginal one))
first the gyro is on the rift, and to be of any use it should be on the robot (to measure balance).
to move a robot you would need some sensor (button, joystick, body suit, compass, and at worst gyro) to indicate at least some direction.
this information will be transformed into commands (go here, go there).
you can eventually use the gyro of the rift to calculate a relative position (from a starting position).
The problem of using the gyro of rift is then your robot will only look forward, and go straight where he is looking for.
so if you try to look left or right the robot will turn left or right.
a good way would be to get a panoramic camera so you can look around with the rift and the decide if you want to go left or right with another control (joystick?)

the only case the gyro of dk2 could be a control for robot would be some kind of flying drone, very agile.(this is already the case for some of them that are controlled by the gyro from smartphones)

You can move a robot with a gyroscope just as easily as you can move one with a stick, a button, or a key on a keyboard. You can move one with your voice, too. You just have to build the bridge between them.

There's absolutely no reason you can't use the gyroscope to control the direction of movement of a robot. nosys70, your statement that it's "impossible" tells us you're like the people who insisted, "Hah, what an idiot! You can't control a car with a wheel! What a dummy!" I would make use of the magnetometer and accelerometer as well, though the gyroscope on its own could provide for some very interesting gesture based control.

Personally, I want to move a pan/tilt stereoscopic camera on a quadcopter or glider with the Oculus Rift. I haven't had a chance to look at the SDK yet, though. I'll be very interested to see what others do with it. I imagine some may have already done this, but I haven't looked yet.

Using the rift to control a robot is a good idea. You should run with it.

everyone here sucks for not clicking on the link I posted ..you know ..it's just a working prototype of what everyone is talking about ..no big

it is just that the signal sent by the gyro of the oculus is way to shaky to drive robot.
that is why it is impossible (yes in therory you can do it, in reality it will not work).
It could eventually for a flying robot, because sudden drift in air are easy (air is forgiving), but try this on wheel...
What you can do is drive some pointer like it is done for the XBox and kinect (you look in some direction and if you stay steady long enough, it takes it like a validation)
That works well in game (like the space demo in oculus) because the pointer is materialized in software only.
With a robot, it is deifferent since there is no reference (gyro are just gyro, there is no absolute positioning).
So if the camera are fixed, the robot turn and and what you see is turning also in the oculus.
it is a kind of remote loop feedback and these are the most difficult to manage.

it is a lot easier to drive a remote camera. because the camera is going nowhere.

@pitsburghjoe wow wow wow!! that mounted on an RC car / quad-copter would be amazing!! I've seen a quad-copter using an Arduino to transfer the Head movements to a gimbal for a live feed video back to the head set (see youtube)... but latency still seems to be an issue, that system you link seems very responsive and very little latency!! how are the two live video feeds synced for direct input to the rift with minimal latency also?

There are a couple older videos on his Vine page https://vine.co/u/1153817920791891968

The official site seams to be this http://www.narvaro3d.com/

http://www.reddit.com/r/oculus/comments ... n_a_robot/

So as some have pointed out, this is already being done. Very much being done, in fact. In some senses there are even worthwhile commercial solutions out there (although some not Oculus-based just yet).

The bigger problem I see -- which you don't think about much until you dive into robotics -- is the incredibly difficult problem of maintaining an (x,y) coordinate on the robot using any of the above sensors. Add in the algorithms to deal with when that goes wrong, there's a lot of complexity. I did a lab in college where we had to make a robot navigate a maze we gave it in digital form to the "cheese" (goal). By far the hardest parts were, for example, say you want to go back and to the right. You do some math and send the following commands to the robot

ROTATE +101.25 degrees
GO FORWARD 12 meters
ROTATE -16.51 degrees

There are small bits of imprecision in the rotating and locomotion mechanisms that throw stuff off, and the errors propagate.

IMO this would be your biggest development hurdle by far! In class most of us used computer vision to compensate -- e.g., if the robot sees a wall, STOP GOING -- and also little heuristics related to friction estimates and approximation stuff (e.g. ROTATE only in 1 degree increments, until passing a certain "point", and then go back 1 degree).