Programmable foot platform - Haptic walker

[petersmc]

Hi Folks,

Palmer has insipired me, so I wanted to start a bit of research on haptic foot platforms. A bit of googling presented these 2 devices, which represent the basics of the idea.
1. Sarcos Biport
2. Gait Master developed by Iwata

I would see something as a mixture of the 2, which is to say; 2 independent platforms that can trace most of the range of movement of each foot whilst keeping the body's center of mass more or less stationary. The theory being that this should allow you to walk (and hopefully run), as well as climb stairs (perhaps ladders), jump and turn in a virtual landscape.

Obviously, the intertial effects will be off, but having said that, this could be a plus in terms of physical effort expended.

Has anyone had any experience with this kind rig? I would be interested if this is a dead end, and if so, for what reasons. Assuming that there is still merrit in looking into this further, then if anyone is keen on teaming up a bit, even if just as a "sounding board", then that would be great.

As background, I trained as a mechanical engineer and have some experience with mechatronics/hydraulics and control theory. Having said that, I work in the financial IT industry now, so I am not yet up on the "off the shelf" products available currently (am thinking mainly electric linear actuators, as these seem to have developed a bit since I was engineering).

[brantlew]

Well it sounds very ambitious - a lot of hardware and control software fusion involved so the best of luck to you. There was a long discussion a few months ago that touched on a number of these type of systems. Most less advanced than your proposal but still an interesting read. With your background you might be able to provide some practical guidance to some of these wild ideas.

http://www.mtbs3d.com/phpBB/viewtopic.php?f=120&t=13784

[bobv5]

I'm resonably good at electronics, but not good enough that I would be prepared to design a system that could dislocate bones if it went wrong. In theory, the electronic control systems wouldn't be hugely complicated. Most likely a more experienced engineer could design that part of the system in a few weeks. Of course, unless he is a member of this forum he would expect to be well paid for his work.

The big problem will be the software. In my opinion, a software guy would have to sort out all the real complexity, and it would be a multi-year project even with decent funding.

I hope I'm wrong. I suspect I am not.

[Fredz]

I agree that the most complicated part seems to be the software. I think the best approach would be to implement a simulator to validate the concept, but it's going to be quite difficult to write.

[cybereality]

Complex, yes, but I don't think impossible. You could probably use an off-the-shelf 3d engine with physics support to do some quick prototyping (for example, with Unity). Most physics engines will allow you to build simple machines and you could use this to test out ideas. For an experienced software developer it would not take several years to write this, it could probably be done in a matter of months (working full-time). At least to build a proof-of-concept, making a consumer-ready product and the testing/QA involved would obviously take longer.

[WiredEarp]

The easiest way IMHO to do this would be to use a pair of Novint Falcons to provide feel/feedback and a passive type rig to do the heavy lifting etc.

[petersmc]

I have thought a lot about this. The idea described is a dead end imo. The main issue is the potential to fall, which defaults to the reguirement of a harness of some kind. For many reasons, i think the harness therefore becomes the primary mounting point, rather than the floor. More to follow in a new thread.

[zalo]

While walking around thinking about this issue, I noticed something:
1. The biggest problem with VR treadmills is that you need to support moving on two axes, not one.
2. When you turn to change direction in a normal walk, you're usually standing on one foot and twisting your body with your ankle.

What if you canceled out the ankle's rotation with a ball bearing? That way, as you walk forward and turn, you'd still be facing forward in the real world. All you'd need to do then is measure the amount of turn, and put the VR-user on a treadmill.

Cue crappy picture of VR-Shoe:

Sherp.GIF

The bearing would measure the spin amount (and when your foot is touching the ground) and translate that to the virtual world. When your foot lifts off of the ground, the sole's sole would then spring back into place (the spring/damper would simulate the inertia of real turning during the turn).

This way it still feels like walking and turning because your senses are fooled by the HMD; you're actually traversing ground space using a natural walking motion. It's not curved like those stupid ball things either.

I suspect that you'll still have to include some sort of alignment-correcting mechanism into the system to keep the user pointing forward on the treadmill. Maybe a HUD indicator? Motorized bearings?

[cybereality]

Interesting concept zalo. I think it could work.

[zalo]

I just realized that bungee cords attached to your belt loops and the side handle bars of the treadmill could stabilize you and keep you safely oriented forward on the treadmill and confident of your next step.

The downsides here are that it prevents crouching and it could feel weird having invisible ropes pulling on your sides.

EDIT: I realized again that since the bottom foot pads will never be turning past 360 degrees, you could use a rotary potentiometer to measure turn, bringing the total cost of the project down to dirt cheap besides the treadmill and positioning sensors (and the HMD and the head tracker (for which the razer hydra is suddenly perfect) and the computer running the whole deal).

EDIT: Another plus of the bungees is that you can use a tension sensor on them to determine when the user is off center (to activate the treadmill to move backwards as you walk forward).

[petersmc]

Yeah, the pivoting issue is a tricky one. Most gait analysis says the pivot and in fact turning is a complex mix of hip rotation and foot pivot. Without some prototyping, i cant say for sure how well this can be solved. I would suggest that as a minimum some potentiomter output at the foot twist could at least rotate your avatar's body, which may be enough. Incorporating hips will be tricky ands adds another level of control complexity.
Wrt bungy feeling - i dont think there is any way around this. One way or another, the body's center of mass must return to a locum. Having said that, this could be a good thing as you wont exert the same effort for avatar movements as in reality. A 3 hour gaming session could be very tiring otherwise. Although, then we would all get olympic fitness . . .

[petersmc]

Some very detailed info on what I was thinking (from p160 if you dont want to read it all):

http://www.ti.rutgers.edu/publications/ ... thesis.pdf

which is to say, the Rutgers Mega–Ankle seems to be the nuts, assuming you can Tim Taylor it for "more power"!

[Okta]

While walking around thinking about this issue, I noticed something:
1. The biggest problem with VR treadmills is that you need to support moving on two axes, not one.
2. When you turn to change direction in a normal walk, you're usually standing on one foot and twisting your body with your ankle.

What if you canceled out the ankle's rotation with a ball bearing? That way, as you walk forward and turn, you'd still be facing forward in the real world. All you'd need to do then is measure the amount of turn, and put the VR-user on a treadmill.

Cue crappy picture of VR-Shoe:

Sherp.GIF

The bearing would measure the spin amount (and when your foot is touching the ground) and translate that to the virtual world. When your foot lifts off of the ground, the sole's sole would then spring back into place (the spring/damper would simulate the inertia of real turning during the turn).

This way it still feels like walking and turning because your senses are fooled by the HMD; you're actually traversing ground space using a natural walking motion. It's not curved like those stupid ball things either.

I suspect that you'll still have to include some sort of alignment-correcting mechanism into the system to keep the user pointing forward on the treadmill. Maybe a HUD indicator? Motorized bearings?

I think this is one of the best ideas so far. As Brent said we have a large thread on this subject and I really like your idea. Treadmill and foot rotation sensors to affect compass facing, with head tracking separate. Sounds like the simplest answer to such a complex problem. I just did some walking to get a feel and i think if you track your right foot rotating outwards to turn left and vice versa it would feel half decent. Any practical suggestions to rig up the foot sensor? Just to add complexity, how about drag you right foot outwards to strafe left ? Toe and heel sensors...

[bobv5]

"For an experienced software developer it would not take several years to write this, it could probably be done in a matter of months (working full-time). At least to build a proof-of-concept, making a consumer-ready product and the testing/QA involved would obviously take longer."

http://en.wikipedia.org/wiki/Ninety-ninety_rule

Proof of concept, months, sure. Code that you would trust your bones to? Even for hobby use I would expect delays.

"The downsides here are that it prevents crouching and it could feel weird having invisible ropes pulling on your sides."

The simulation software could be made in a way that do not require crouching.

"EDIT: I realized again that since the bottom foot pads will never be turning past 360 degrees, you could use a rotary potentiometer to measure turn, bringing the total cost of the project down to dirt cheap besides the treadmill and positioning sensors (and the HMD and the head tracker (for which the razer hydra is suddenly perfect) and the computer running the whole deal)."

Cheap potentiometers suck poo poo. Good potentiometers cost more than optical rotation encoders, and they are still less reliable and less accurate. Two or three degrees might be significant when it is where a robot should force your legs to be.


"A 3 hour gaming session could be very tiring otherwise. Although, then we would all get olympic fitness . . ."
First of all, 3 hours in a simulated warzone should be tiring. I suspect using imersive VR gear, it would be tiring even with minimal haptic feedback. Second, the physical limts of the user would relax the requirments of the hardware. Make the game occur at walking pace, and you have probably halved the cost of the hardware.

[petersmc]

OK folks - I feel a plan coming together. Consider this:

http://www.theverge.com/2012/1/2/267755 ... -treadmill

Remove the treadmill and add freeclimbing harness to support the body, and then some lower body exoskeleton that can allow you to support your bodyweight on your legs when standing. Add some dampners to offer some resistance for a walking motion. Cheap and cheerful for a start.

It also seems that there are a number of kintect hacks that already allow limited avatar control.

Anyone know Taylor Veltrop?

[petersmc]

Just before I get hit too much about the latency of the kinect, check this out:

http://solidsmack.com/3d-cad-technology ... -to-shame/

Seems we are in the sweet spot of tech convergence to get very interesting affordable VR solutions.