Envisioning a DIY CyberCarpet for Omnidirectional Motion

[cadcoke5]

Given the amount of work and money in such a system, I wonder if there is a way to test it to some extent. Might you simply go somewhere that has a gravity roller system and ask to walk on it? I would only try very small operations, so that the chain of authority is very short... ideally only one person and someone who already knows you. Otherwise, they may tend to think about liability issues instead of how cool the idea might be.

I think the test would establish if you are comfortable walking over rollers at the diameter you are building them.

Another test method is to just draw the arrangement of pipes life-sized onto paper and place it on the floor. This will help you to visualize using it. Such exercises can help you to imagine it in use more concretely than if you are drawing on your desk top.

Since we are brainstorming, here an idea regarding a cheap roller system. Imagine vacuum forming a sheet of plastic over a bunch of pipe segments. Then, remove the plastic and drill or punch small holes along each groove formed. Assemble this to form an air-hockey table arrangement so that each pipe can be placed in a groove and it would be free to roll because of the air supply. If you have enough air holes, then you may be able to get away with plastic pipe instead of metal.

You would still want something to keep the pipe from rubbing on each end, but that doesn't need to be much. Perhaps a plastic ball sized to too big to fit completely inside can be glued into each end.

The drive mechanism might be accommodated by placing a strip of wood over the pipe before the plastic is formed over it. Then, that wood would leave behind a groove that a drive belt can run inside.

Joe Dunfee

[PalmerTech]

No time to address everyone else, but as far as trying walking on a roller system:

Most gravity roller systems are actually pretty widely spaced, and they are almost totally frictionless. Walking on one would not tell me much.

I do plan on chalking out the size, and I also plan on spending a little bit of money building a single, non motorized platform. If that goes well, I will develop it further.

[android78]

I haven't read every post here, but I think the only really economically viable option is the roller shoes.
The issues with this seem to mainly be related to four things:
1. using normal roller blades, you can only go forwards/backwards, not turn or strafe
2. It is difficult to limit the rolling so you don't just fall over.
3. How do you track the movement?
4. How to stay in the one spot?

So, my idea for these is:
1. Use some regular roller-skates (four wheels with two at front, and two at the back) except replace the wheels with these:
http://www.acroname.com/robotics/parts/ ... LER-3.html" onclick="window.open(this.href);return false;
2. If you have seen the 'soft closing' cupboard/draw mechanisms, they get a nice stiff action by just using a heavy grease. Most roller-skates use a fairly light grease to make them as friction-less as possible, so just washing the bearings and replacing the grease should get a nice and consistent friction.
3. The main Idea I have here is to use the sensor out of a regular optical mouse, mounted with a spring that will press it against the ground lightly when you put your foot down. The movement tracked by the mouse will be opposite to what you want as input though. Also, you need two mice to track both feet, which will make it difficult to write a driver for.
The other option is to track the wheels, but this would be more difficult.
4. I believe a harness that just goes around your waist with some elastic attached to four (or more) points on a heavy frame is all that would be needed.

[android78]

I've too many other projects on the go, but if anyone was going to implement the omni-shoe concept, the following might help:
http://www.ebay.com.au/itm/Omniwheel-Fl ... 239wt_1139

[cybereality]

So, my idea for these is:
1. Use some regular roller-skates (four wheels with two at front, and two at the back) except replace the wheels with these:
http://www.acroname.com/robotics/parts/ ... LER-3.html" onclick="window.open(this.href);return false;

OK! Now we are talking! This looks very interesting. I wonder how hard it would be to retro-fit those wheels onto a pair of roller-skates? That certainly seems like the most affordable suggestion so far. Those wheels were like $7 each, for 4x thats $28. Maybe $100 for a pair of skates. And the tracking could use multiple off-the-shelf equipment (Wiimote, Kinect, etc.) and only add maybe another $100 to the cost (not counting the computer and all the other stuff). I'd like to explore this further.

[Okta]

Those wheels don't look like they can bare much weight. That aside, if you make some skates from them, what then? Doesn't that only leave you with the option of suspending your full weight in a harness and just have you feet touching the ground for walking motions and still having an issue with turning?

[android78]

I don't believe you need to be fully suspended in a harness. the harness prevents you moving forward, backward, left, right and turning. your full weight would be on the wheels. you'd have to test the wheels for weight, but think they would be good enough for testing. There are similar products that are all metal that could be used in a final product if found satisfactory.
The sensors on the shoes would monitor the floor position relative to your shoe, which will be the exact opposite of your movement relative to the ground if walking normally. This would probably need two sensors per shoe to get turning as well as forward/backward, left/right.

I think an ideal setup would actually use only 3 wheels per shoe. one in front of toe pointing left/right and two just behind the heel so they are at 60 degrees to the front one. Ideally these would also have powerful supper motors attached to keep the wheels all moving at the same speed though.

[brantlew]

Well this is interesting. One of the more practical ideas that came out of this thread was just simply using a Kinect to track gestures and stationary walking. Well this guy went ahead and did just that. Not as exotic as many of the other ideas put up, but you gotta admit this solution is simple, cheap, requires minimal space, and achieves many of the goals of a natural motion system. So is this more immersive than a button controller? You be the judge...

[youtube-hd]http://www.youtube.com/watch?v=7vcGqha6xJ0[/youtube-hd]


http://zookal.com/blog/2012/03/how-to-p ... l-reality/

[Synexious]

Only interesting if it works in 360 degrees.

[PalmerTech]

I am sure it could, the Kinect is pretty good at handling rotation.

[WiredEarp]

Looks better than I expected...

[pierreye]

Actually he is using TrackIR 5 for head tracking so basically it can't handle 360 degree. You should look at the multiple take at the end of the clip which is quite funny.

[BillRoeske]

Migrating the relevant parts of this post over from the independent head and weapon tracking thread (along with a follow-up to brantlew's reply):

Back in college (2003, SUNY Buffalo), a friend and I hacked together a setup with one 6DOF tracker on the HMD and another attached to the user's belt, centered in back. His or her forward vector for movement was whatever direction their hips were facing, independent of where the HMD was looking. We never got to the point of adding a weapon, but an independently tracked wand would have filled the role, there.

For small movement corrections, the user could just move around by one or two steps. For longer journeys, the user could walk forward by standing upright and walking in place, sidestep by leaning to the side and walking in place, and backpedal by leaning back slightly and walking in place. We constructed a pretty Frankenstein set of two-part "sandals" that would strap under the user's shoes. They had contact sensors in the ball and heel areas, which were wired up to the button sensors on an old Thrustmaster wireless gamepad. We did some rudimentary gait detection based on the cadence of the foot switches being turned on and off when someone was walking.

We never got to the point of refining the setup, but it worked pretty okay for what it was. By that, I mean most users didn't almost fall flat on their face as soon as they tried moving around. For those that did, I blame the fact that we built the foot sandals out of 1/8" plate aluminum.



I have some fabric ones sewn together, but they never were put to use.



For a cheap DIY home set-up, it wouldn't surprise me if a cell phone with a magentometer and accelerometer (pretty much all smart phones) could replace the hip sensor and also help detect gait. Not sure how you'd replicate detecting lean without trackers with absolute position, though.

[BillRoeske]

It's hard to pass a definitive judgement on how successful a rig like ours could be, mostly because of the total lack of time we had to refine it. The whole thing was thrown together by my friend and I during the span of a semester as a means to an end while working on our actual project (don't laugh ) for an undergraduate course.

Here are the big things that I remember:

• It was intimidating, especially for first-time users. It took around three minutes of swarming around someone from head-to-toe to get them fully suited. The aluminum foot plates and black box attached to the hip were especially foreign-feeling. Having a good bedside manner and briefly explaining what each part did as we went to put it on helped the whole process. Still, seeing the clunky setup and having a handler managing the umbilical of power, video, and tracker cords when someone was playing made bystanders at the show reluctant to try it.
• Walking felt weird. Almost everyone knows exactly what walking feels like, and walking in place doesn't feel like walking. For example, movement didn't start until after the user had put their foot down. We tried a couple of experiments with predicatively starting travel when a foot was raised, but our code guessed incorrectly more often than not. Simple things like shifting your weight could cause a step, which felt very wrong. Granted, our movement was hacked in on top of Quake 2 physics, so it was functionally the equivalent of having a mouse wrapper for HMD orientation.
• People generally look where they want to walk, so having the viewing direction separate from movement direction wasn't a huge revelation. In fact, it confused the hell out of some people when they tried do do it after enduring us explain how cool of a feature it was.

Because we were pressed for time, a lot of what made it into "production" were first-run ideas. Modern equipment and more time spent on the software could certainly help, but I think that I have to agree with Palmer that nothing will replace actually being able to physically walk around. There were times when it worked okay and felt pretty good, but I'm not sure if it could ever truly climb out of the uncanny valley of movement.

That said, I still have the slip-on fabric version of those boots hanging around. Palmer and John have certainly been inspiring with the RIFT and Doom 3 work, respectively. It's certainly a tempting idea to dust everything off an give it another shot, particularly once I have my hands on a HMD again.

[brantlew]

Almost everyone knows exactly what walking feels like, and walking in place doesn't feel like walking.

Yes, that's what is so difficult about motion control. It is so hard to reproduce the correct motion and weight placement with contraptions.

People generally look where they want to walk, so having the viewing direction separate from movement direction wasn't a huge revelation

I think it depends a bit on the context. When people are purposely walking towards a goal they keep their head forward and turn their body. When exploring however, I think there is a tendency to walk slower and glance around independently a lot more. Although the angle is still confined to maybe +-45 degrees. People almost never walk with their head turned completely sideways.

Technology has improved so much in the last ten years, I think you would have fun dusting off this idea again. There are so many ways to implement this concept nowadays. Thanks for sharing. I'm a sucker for this sort of conversation since motion control is my own current area of research

http://www.mtbs3d.com/phpBB/viewtopic.p ... =60#p73183

[cybereality]

@BillRoeske: Thanks for posting this, sounds like an interesting project. Seems like maybe it didn't work 100% to your liking, but I think the idea is sound. Its mostly a matter of pairing the right hardware with good software. I feel like we have the proper technology to it right, just maybe no one has cobbled everything together.

[BillRoeske]

Thanks for the encouragement, guys. You've got me at least thinking about it again, which is step one (no pun intended). I'd be a liar if I said that I didn't spend at least part of last night walking around my house like an idiot, trying to think carefully about the feel of it all. I have a very supportive wife.

One thing to consider in the short term may be how to gather some meaningful data about walking. Does a long gap between when a foot is raised and then lowered mean a large stride or a slow sneaking step? People tend to vary their height differently when moving quickly vs. slowly, so perhaps there is some reasonable correlation of time between footfalls and amount of head bob as an indicator of intended distance to travel.

I'll think about it some more. If it goes anywhere, I'll start up a project thread for it once I clear up some of my other obligations.

Oh, and brant: Looks like you've made some really great progress there! Consider me a regular lurker of your thread now.

[android78]

I just had another light bulb idea for a large park kind of setup. If you had a string that went from one foot to the other that was attached to a spring at one end, and you could get absolute length of the string, combine this with pressure sensors in the shoes and a potentiometer to measure the angle that your foot is pointing and you should be able to work out any movement and direction change very cheaply.

[cybereality]

@android78: That could work. The GameTrac used a similar method. It would give you one foots relative position from the other (though you still wouldn't know where the first foot was). Maybe that would be enough to make some assumptions about the desired motion. With the sensors on the foot you would at least know which one was on the ground. So it make be enough. Interesting.

[android78]

I don't think you need absolute position that you are starting. In a game, all movements are relative to the current position, so the relative movement could be calculated based on half the distance between your feet from the foot that is considered down. So if your left foot is down, then as you move your right foot forward, your position would move half way between your two feet. Then when you put that foot down, the position should be pretty accurate relative to where you were.
As experienced by Brant with the guided walking, I don't think any small errors in the position tracking would be noticeable. The problem is obviously that you would need a huge park, or combine it with GPS (or Brant's tracker) that could guide you when you start to approach the boundary of your area.

[android78]

@Cyber, I did a quick search and can't find anything about gametrac. Any links to this?

[cybereality]

Sorry, it was GameTrak with a "k":
http://cb.nowan.net/blog/2006/09/25/gam ... mpression/

[android78]

Thanks for the link. It looks similar to what I've in my head, but for hands. I think you're more likely to get the strings tangled with your hands though, especially with the two strings.

[zalo]

That's kind of a nice idea (could be done with a rotary encoder/spring combo for the distance/tension sensor).

My only concern is with the relative rotations of the ankles, you could get weird readings. Assuming you're not on a bidirectional treadmill, whenever you turn your feet the potentiometers will feel it and throw off the whole thing. At that point, it will be difficult to get relative positions. Maybe you could solve it somewhat by having one foot be the "reference foot" that it assumes is always pointing forward.

[android78]

The "reference foot" is determined by the pressure sensors. The one with most pressure is the one that is considered stationary. It could be a problem if you swivel on your heels though, but I think that's a hard one no matter which solution you use. In the same way that the movement forward is half way between your feet, the angle would be half way between the angles of your feet (relative to your 'down' foot). I'm not sure what problem the bidirectional treadmill solves in this situation. I imagine that having omni skates that keep you in the same position would be ideal for use with this, but if you're going to those lengths, then you can just get position from wheel counters.

[brantlew]

I don't think you need absolute position that you are starting. In a game, all movements are relative to the current position, so the relative movement could be calculated based on half the distance between your feet from the foot that is considered down. So if your left foot is down, then as you move your right foot forward, your position would move half way between your two feet. Then when you put that foot down, the position should be pretty accurate relative to where you were.
As experienced by Brant with the guided walking, I don't think any small errors in the position tracking would be noticeable. The problem is obviously that you would need a huge park, or combine it with GPS (or Brant's tracker) that could guide you when you start to approach the boundary of your area.

This concept occurred to me as well a few weeks ago when I saw Chriky's post about a foot mounted Hydra. Using magnetic trackers on the feet relative to the torso seems like an even simpler method and would adapt extremely well to the types of things I have been working on. I bought a Hydra about 2 days afterwards with the plan of exploring this idea thoroughly as soon as I can get some free time.

[Chriky]

Awesome I was hoping someone would take that a bit further, it looked like it had potential, I'm tweaking my HMD and using the Hydra for head and gun tracking (with stationary base) at the moment, so had to dismantle the Hydra-shoes setup.

One cool thing about that kit is it would also give you crouch (feet close to hips) and jump (both feet not touching) controls.

[FingerFlinger]

This concept occurred to me as well a few weeks ago when I saw Chriky's post about a foot mounted Hydra. Using magnetic trackers on the feet relative to the torso seems like an even simpler method and would adapt extremely well to the types of things I have been working on. I bought a Hydra about 2 days afterwards with the plan of exploring this idea thoroughly as soon as I can get some free time.

Ha, does anybody on this forum even have any free-time left?

Android's idea definitely seems worth exploration, although obviously, when using it in a park, there will likely be tripping hazards with the cord.

[JohnnyGrant]

Hi everybody i'm new to this forum and Just wanted to say hi!

http://m.youtube.com/watch?v=WHfUpI66tfo

How about building a manual version of this? From what I've seen so far, this looks like the most natural walk. would it be
possible to track the leg movement and translate it into movement? Maybe with the hydra? Greetings:)

[brantlew]

I think the Kinect SDK would be simplest for prototyping but the Hydra would probably work also and be better of quality.

[TilliK]

Hi guys! I started reading trough the whole topic, I am really frustrated that there is no "home in your living room" solution yet.

there is some good "big space" options like brantlew's friispace looks totally awsome, but not so practical, You need a whole lot of space and here in sweden we don't have the best weather ^^
and we have the big "hamsterball" also not to practical to have in your living room, and not to talk about the ODT like the one used in BF3 simulator by gadget show. Then we have the wizdish, Good but dont register side to side movment or backing up.

So at page 9 I saw some youtube videos linked by Fredz,

this video was really interesting to me:
[youtube-hd]http://www.youtube.com/watch?v=n3U8enNlhqM&feature=plcp[/youtube-hd]

It looks horrible but has some good things to. Now imagine a better floor with omni directional ballbearings like the one palmer showed at page 4
http://www.vxb.com/page/bearings/PROD/B ... ts/Kit7764 or similar

and some of you talked about suspending the player in a harness. So combine this contraption with something like THIS!
[youtube]http://www.youtube.com/watch?v=_ihooeiAdX8[/youtube]

so now you are suspended so you are just touching the floor and can move freely around (the harness lies on bearings similar to how a bicycle or a motorbikes handlebar works, therefor moves with you) and when you move you are "pushing" forward just a bit and that movement is somehow registered as a analog signal.

It's really hard for me to explain my thoughts due to my limitations in English, But I will try to make some sketches tomorrow! but i hope this gives a basic understanding how I'm imagining this!

Best regards
TilliK

[cybereality]

@TilliK: Hey man! I am very much leaning toward something similar. Something involving a pod and a harness and some sort of omni-directional ball-bearings. I'm sure you can combine that stuff into a workable setup. Somehow...

[TilliK]

@cybereality, Yeah, We need to get away from tracking the Surface we move on and start tracking the player somehow! That's my belief anyway.

[virror]

Hello, im new to this forum but really interested in making player motion possible in VR.
Excluding the "walk in place" solutions i think the roller shoes are best alternatives. Im thinking about a solution were you only track the forward/backward motion and handle the rotation with some external tracker like the rift. A plate with small 1-2cm wheels and a simple opto sensor (like in the old wheeled mice) to measure the speed of the movement. To keep the player still you would have a slight concave surface to walk on. With a little bit of wheel friction and just the right angle on the base plate i think it should be possible to walk without to much feeling of "walking on ice". Tho total cost of this could probably be kept to max 300$, probably cheaper. Biggest issue here is to find that perfect concave surface to have as a base. Any thoughts?

[brantlew]

@virror: Welcome to the forums. Best of luck with the skate idea. Lots of people have talked about it but nobody has followed through with it yet.

I do think friction is going to be tricky. With a concave surface you are going to have varying levels of force. At the steep part the skates will need more friction to resist rolling than they will at the bottom. So for a continuous walking stride you might need some sort of motorized wheels with continuously variable force to compensate for the shape of the well.

[MSat]

Why not simplify the system by using a low friction surface like the WizDish? Requires no moving parts other than your feet

[virror]

To be honest, wizdish looks really strange and not like real motion at all : p

[MSat]

To be honest, wizdish looks really strange and not like real motion at all : p

Right, but just like any low friction system (roller floors/shoes/etc) you ideally want something that holds your body in place so that you can walk more naturally.

[android78]

I wonder if applying a high viscosity fluid to the wizdish could work to give some friction. You would probably have to actually make a much larger wizdish that end up much steeper at the edges for this to work and prevent you walking off the edge when you walk fast though.

[zalo]

What, like a cornstarch and water layer? Could get messy.

Look up paramagnetism with neodymium magnets and aluminum. It's a strictly resistive force, and if you have the magnets in your shoes, you can vary resistance by the thickness of the aluminum plate in the wizdish.