Slashdot Mirror
The Dismounted Soldier Problem
Pilchie continues:
"Some Ideas
These are some of the ideas that I have heard, and some of their drawbacks:
Don't walk. Use some other input system (joystick, whatever) to move the character. Problem: People get lost if they don't actually walk. See here.
2D mesh of bearings. This is probably the best idea so far, the idea is to have a 2d mesh of bearing that can rotate in any direction, each with a sensor to determine which direction it is rotating in. The drawback is the difficulty in sensing the rotation of the beads, as well as allowing them to move in any direction. Plus safety(things getting caught between the beads).
Spheres. This idea involves placing a person either inside or outside of a spere and letting the whole sphere rotate in an arbitrary direction (sort of like a mouse ball). The problem is making a sphere big enough, with little enough friction that it can actually be moved by a normal walking action.
Mime Walking. This involves teaching people to walk in a special manner to allow the system to know they are walking. See this again. Problem is that it is fake.
Large room/visual tricks. Put someone in a large room and let them walk where they want. When they start to approach a wall, rotate the picture enough that they compensate, but not enough that they realize it is rotating. Problem: obiously you can't be sure they won't hit a wall at some point.
Well what are your ideas on the issue?"
Hypothetically, it would work like the logitech "marble" trackballs. The balls would be speckled, with large contrast between the ball color and the speckle color. Optical sensors would then track the movement of the balls, and I would assume, average out the direction and the speed of the movement among all the balls.
One problem would be after-spin. When you spin a trackball, it will continue to spin for a second or two after you let go, depending on how fast you would spin them. Maybe there would be more friction imposed on the bearing mesh. It's an interesting problem.
What ever happened to using a neural signal inhibitor to pick up the walking signals in the spinal collumn nerve cells and translating them into computer instructions?
We have pretty good beta of it going here at [echelon system autocensoring error 431 - no replacement known for this forbidden term] based on stuff the aliens bought left us. And it seems you can use it for a full 2 weeks before you loose your physical motoring skills completely!
The device also has great application within torture and teledildonics (what can I say, we don't get that many woman here at [echelon system autocensoring error 431 - no replacement known for this forbidden term]).
-
We cannot reason ourselves out of our basic irrationality. All we can do is learn the art of being irrational in a reasonable way.
There has been a lot of development in the squid field over the last few years (squids are a type of minute voltage detector), and while it takes a little training (ususaly about 10/15 minutes of practice to get navigable, which is much less than the time it took most of us to learn the nintindo controler) sqid based controlers have been under development for a few years. The idea is you only have to THINK about moving a muscle, and its voltage changes, but it dosn't actually move until you reach a treshold value. With this tech, there are all kinds of fun ways to solve the walking problem. Personaly, I would put squids in an input glove, and have it respond to *pushes*.
-Crutcher
-- Crutcher --
#include <disclaimer.h>
How about this: two robot arms, with shoes where the grabber would normally be. Person stands in the shoes. The arms would controlled so that they move freely when the feet are above "ground", and lock up when they hit "ground level".
:-)
This would have the advantage of also working on stairs or gradients, or even wading through mud.
I have seen something similar in arcade games simulating skiing. These only had one degree of freedom for each foot though. For this to work you would need six.
Of course, things could get unpleasant if the machine crashed. Want to risk being a human wishbone anyone?
Ok. This might be *slightly* off topic, but If it's a great VR experience you are after you shuld try lucid dreaming (basicly beeing aware that you are dreaming, when you are dreaming, and then do whatever you like in the dream).
No problem with walking, eating, flying, doing magic etc...
Well it can be a bit hard to learn of course, but it's well worth it
There is one core problem here: how to maintain walking motion in a restricted space.
User is suspended in a harness, legs dangling, arms free. This can be comfy if done properly. Manipulation controls, gloves etc can be used for arms. On each foot is a position sensor. A certain line just above the maximum reach of the users feet is designated as 'ground'. He moves his feet in normal walking motions, and the positions of the feet in relation to the ground plane are interpreted:
Anyway - just some ideas. I'm very interested in such issues, so go ahead and email me.
As I see it the general problem here is the inner-ear based sense of acceleration. We feel it when walking, riding an elevator, falling, turning a corner in a car, etc. When the visual sense of place and the inner-ear sense of acceleration are in conflict, most people feel disoriented, sometimes to the point of nausea or vomiting.
In constant-speed walking, the acceleration perceived is largely up-down, except during the first and last steps. Perhaps an appartus with a wide range of movement could duplicate both the up-down acceleration (a simple matter of "bobbing") and the initial and final accelerations, each of which would be followed by a gradual (enough not to be noticed) acceleration in the opposite direction, to ensure an eventual return to center. My gut feeling estimate is that to be "believable" a machine of this design would have to have a diameter of about ten meters. Big and expensive, but perhaps doable.
rat
Pretty funny, but not totally far from the best solution. We're actually reasonably close to being able to decifer nerve impulses, so controling walking via a tap into the spine isn't such a pipe dream. But there are some serious drawbacks regardless. 1) Pattern generators. Walking is a standard "program" the nervous system can "run", meaning it CAN do so without sensory input, and we can measure it. The problem is, PGs also use various sensory input to do "error correction," which means we could get all sorts of addative errors and confusing screwups. 2) Without actually walking, how are we going to make it "feel" like walking? Reproducing sensory feelings is way harder than measuring standardized spine output. It would feel extremely uncomfortable to "walk" without the proprioceptive feeling of walking. 3) Related problem- how are we going to shut down the muscles without shutting down the proprioceptive nerves in the muscles, which would also exacerbate the problems of 1 and 2
I decided to respond to this at the top level because I have seen numerous people state things about it, and I wanted to address all of you. Contrary to popular beleif, the inner ear problem is not much of a problem at all. If you provide an immersive enough visual environment, the human brain will compensate for not receiving the acceleration inputs for moderate accelerations, such as those involved in walking, etc. This is one of the things that the group I worked with researched. In most applications, true acceleration inputs are only required to differentiate self motion from external motion. As an example of this, in the helicopter simulator I worked on, a motion platform was required for the pilot to be able to determine which motions were of the helicopter, and which of the ship. However, for someone standing on the deck of the ship, not watching another moving object, there is no need for a motion platform. Yes it makes it more realistic, but it is sufficient for training purposes to provide the visual stimulation. In fact most of the time when we were testing we didn't have the ship moving, and therefore didn't put the motion platform on, because it was unnecessary.
>~~~~~~~~~~~~~~~~
>~~~~~~~~~~~~~~~~
Pilchie
I work in the modeling and simulation domain. One team in our local organization is working on Individual Combatants (the more general term for what we used to call Dismounted Infantry). They participated in some exercises to test and demonstrate new IC technologies. One of the coolest things I saw was an omni-directional treadmill. It was really cool. It even had some mechanical devices attached to the user's torso that could simulate the forces involved in going up or down a hill. You can find more about it here.
Pilchie--
i efsp59.html) Cybernet, Inc., under contract for the U.S. Army, has developed a "foot-haptic" (i.e. tactilely responsive to the feet) system for modeling locomotion in VR. Patent pending. Found via a search on "kinesthetic analysis".
You're approaching it from the wrong direction. "Think outside the box", dude.
Don't build a user input device. Model the leg's motion directly!
Here's what I see you trying to do now: through a "natural" walking process, you want to have the walker's feet create a change in some input device (much as a hand creates a change in a mouse or keyboard or joystic). Then you measure the change in the input device and convert that to a model of motion.
But you're sticking in an extra (and probably unnecessary) layer of abstraction: Information about legs' movement CONVERTED TO Information about input device movement CONVERTED TO virtual model of legs' movement.
Input devices makes sense in modeling vehicular movement. For example, in modeling driving a car, you don't really care about the motion of the hand itself; you care about the motion of the steering wheel. (The fact that a human hand is guiding the steering wheel is irrelevant. From the standpoint of modeling the vehicular motion, it could just as easily be a monkey or a computer moving the steering wheel for all you care.)
A steering wheel becomes the input device to control a virtual vehicle, because in the real world it ALREADY IS the input device used to control the vehicle.
Input devices don't make sense in modeling movement of the human body. In modeling walking, you care about the motion of the feet -- their direction, the length of the gait, etc.
In the real world, what is the input device that is used to control the motion of the feet? THE LEGS!!!
Try this: Information about legs' movement CONVERTED TO virtual model of legs' movement.
Gets rid of the input device.
My idea: Instead of trying to measure the rotation of hundreds of spinning bearings, measure the motion of the legs in relation to a fixed point. (Actually, all you probably really need to know is the location of the footfall, so you probably only need to measure a couple points on the foot in relation to that fixed point.)
How? The simplest example I can think of off the top of my head is kinesthetic analysis of athletes, like for golf swing analysis. The golfer and the club are covered with a series of white dots, typically at each joint, and is digitally filmed swinging the club. A computer analyzes the digital video, recognizes the dots, and uses them to construct a wire frame figure in a virtual space that can then be analyzed. It's used not just in golf, but in many sports as well. It's also used in dance!! I've heard of several dance projects out there that are attempting to use the human body itself as the interface to control or model a virtual dancer.
If you want to model the motion of a jet fighter, who do you go talk to? An aeronautical engineer! Why? Because that's the person who is going to know most about how a jet fighter moves and how that motion is controlled.
If you want to model the motion of a human body, who do you go talk to? Either an athlete or a dancer (or someone who studies athletes or dancers)! Why? Because they're the people who are going to know the most about how a human body moves and how that motion is controlled.
So maybe you have dots on their feet and a camera trained on their feet, a computer watching the dots, crunching the numbers to model the motion of the feet, and passing their location onto the VR modeling system. Maybe you have a super accurate GPS system (or just some kind of very localized version of a positioning system) that sends back the specific location of the feet to the VR modeling system. If you know the how the feet's movement is changing in relation to a fixed point, then you know the direction of the walk, the speed of the gait, etc.
You still need to decide how the subject (the real subject, not the VR avatar) interacts with the real environment while immersed. Probably the *easiest* way is to give the subject plenty of room to maneuver. Put them in an airplane hanger with a VR headset on. Let them move around to their heart's content in the hanger, model the motion of their feet (and hands! and head direction!) and build the virtual world around them as appropriate, based on that motion. It would look funny from the outside -- a couple of soldiers in headsets wandering around a hanger. But from inside the VR they might be in downtown Beirut or wherever.
With that plan, you *will* run into physical limits -- it would be possible for them to bump into the hanger wall. If you don't need an infinite virtual space, then big deal. If you do . . . eh, you might be able to do the "large room/visual tricks" option to make them change direction. Howver, that's probably too complex, I think.
If you need an infinite VR space, you might need to put them on that 2D mesh of ball bearings, so they can "walk without going anyplace." (You just have to make sure there's enough ball bearing friction that the subject doesn't fall on his ass!)
You don't need to measure the motion of any of the ball bearings. Just ignore the ball bearings. The ball bearings are just the foot's medium of motion. If you want to model head motion or hand motion, you don't try to track the displacement of all the air molecules around the head or hand! No, you track the motion of the head and hand itself in relation to a fixed point. Then why try to track the displacement of the "ground" (aka ball bearings) beneath the feet!!?? If you know how the feet are moving in relation to a fixed point, then you can use that information to model the virtual motion.
TRACK THE FEET, DUDE!
--
fixion
fixion@yahoo.com
P.S. Two minutes searching the Web netted me these links:
Peak Performance, Inc. (http://www.peakperform.com/) has a product called Mocap that captures 2D & 3D motion coordinates with real-time optical sensors. Found via a search for sports technology and biomechanics.
VNSIII (http://www.interlog.com/~drokeby/vnsII.html) allows you to respond and analyze motion information captured in real-time. Found via a search for "dance and technology."
and the kicker:
Whole Body Kinesthetic Displays (http://www.cybernet.com/rnd/contracts/contractbr
Translation: US Army beat ya Canadians to it, dude!
And that's just, like, the first three promising ones I came across!
Does anyone remember the baby-walkers?
Perhaps you could have the person supported on an item that could be dragged back to the center of the room without their knowledge. That way they can walk in any direction and never hit a wall. When they come near a real wall, the user is visually prompted to raise up his/her feet so that the mechanism can return to the base.
There are a few disadvantages though, like having to build the thing to hide the acceleration back to base. Also, the lifting of the feet could be troublesome.
-B