Hacker News Reader

Comment by jameshart | Hacker News Reader

This seems like a poor fit for VR motion simulation - actual movement through space is generally unnecessary, you just need to gimbal the platform to simulate accelerations. Actual linear acceleration is of limited value in simulation because you can't maintain that acceleration indefinitely - you run out of room and have to apply an opposite acceleration, which can conflict with the simulation needs.

On the other hand this looks more useful for something like movie making, to throw actors around in front of a greenscreen.

View on HN

I've been working with full motion simulators for the past two years (I even rode that simulator last week at MPI) and I must say that this is plain wrong. This is actually a great fit for VR simulation (see below):

> Actual linear acceleration is of limited value in simulation because...

Linear acceleration is immensely useful for sudden accelerations. Image a sudden linear acceleration (e.g. a racing car accelerating). You cannot recreate this acceleration by suddenly rotating. The angular accelerations would ruin the experience. Rather, you make a linear movement, and THEN you rotate the simulator (if you need to sustain the acceleration).

> you run out of room and have to apply an opposite acceleration, which can conflict with the simulation needs.

wrong again. If you move the simulator at slow speed, you can actually come back to the optimal position and the subject does not notice it (washout filter [1])

[1] https://en.wikipedia.org/wiki/Motion_simulator#Implementatio...

That's fair; turning motion into lean gives you best of both worlds - and there is, at least, with this the possibility of up/down accelerations, which are the hardest to pull off convincingly.

Would adding rotation to the seat allow for sustaining the acceleration?

The way I understand what he said (see quote below), yes it would.

> you make a linear movement, and THEN you rotate the simulator (if you need to sustain the acceleration).

I have zero experience with this though, I'm merely interpreting and copying what ghgr said.

That's right, and this system seems perfectly capable of that.

> Rather, you make a linear movement, and THEN you rotate the simulator (if you need to sustain the acceleration).

Am I understanding this wrong or are you saying long accelerations can be achieved by accelerating, turn, accelerating, .... etc? Wouldn't the person wearing VR headset feel the turn?

Another way of saying this: we're good at sensing the delta of acceleration, not the absolute acceleration. Also, we tend not to notice small DC values of either acceleration or rotation, so you can manipulate the user's perception.

Someone published a similar hack for full-room VR, showing you can 'bend' someone's trajectory, to create the false sense that they're walking straight when in fact they're curving. In a large enough space you could simulate an infinite hallway (by having the user walk in a circle).

Sorry, I tried but can't dig up the link right now... but it's out there somewhere.

http://cb.nowan.net/blog/2008/12/02/redirected-walking-playi...

I think you were searching for that. It seems it also comes from the Max Plank institute.

That's the one. YourGoogleFu >> myGoogleFu.

Funny, I think Star Trek TNG described how the holodeck works by saying that it tricks the user not to walk into the walls. The visual feedback of slowly rotating hallway would definitely overdrive the inner ear. Too much, though, and you're going to get sick from it.

Let's say I want to simulate acceleration going forward. First, there's the burst of forward motion provided by the cable rig, and that can be > 1G. By also gradually flipping the person onto their back you can simulate continued acceleration of 1G because lying on your back feels the same as 1G of acceleration forward. The VR rig would still show you accelerating forward. I am guessing that if the rig also slowly moved back to its original starting point the user wouldn't notice.

Thanks for explaining, I kind of understand it now, but what if the acceleration is > 2G, then gravity won't be enough to simulate it, and I'm still a little skeptical that the user won't feel the rotating and "slowly move back" to original position, but I've never used a VR headset so I might be completely wrong here

Your body's not that good at measuring acceleration. Sure, it'll be hard to convince your brain that blood's pooling in your feet, but being leaned way back will provide a pretty convincing >1g forward acceleration feeling if the visuals are right.

This is really neat! I'm really curious what the motion would look like from outside the room, and what the limits of a simulation would be using this technique...

It seems like the sort of fun project that would be cool to simulate in a graphic engine -- pairing the playing of certain games with a 3D model of the simulation room :)

I think hes saying you accelerate forward then tilt back to let gravity help out.

Gimbals are limited to ||acceleration||=1g. I guess there are some elements you miss from that, like when a roller coaster starts to fall (you experience <1g), or when a car takes off (you experience >1g). Here you can provide a bit of both and also gimbal action. There's also rotation effects (which I don't know if are felt) that can be mitigated too and better jerk simulation (not only acceleration, but variation of acceleration).

Actually if your whole simulation has ||aceleration||>1g you should be able to do it perfectly (save for jerk/rotation effects) with some large centrifuge: the rate of rotation of the large centrifuge dictates the "perceived gravity" and gimbals dictate the "perceived direction". In this sense it would be perfect if you could get a centrifuge with a gimbaled end in a 0g environment.

Hey NASA, I have this one little idea for the ISS... :)

NASA seems to think different. http://www.aviationsystemsdivision.arc.nasa.gov/facilities/v... https://www.youtube.com/watch?v=fhu_UldlP_E (in motion...)

For flight, the problem is that the only acceleration that really matters is variation in vertical acceleration. Linear acceleration can briefly simulate that, but it can't be sustained. That's a fundamental problem that can only be dealt with by creating that kind of massive vertical acceleration environment - even that will be limited in terms of how long it can simulate sustained > or < 1g. You can create sustained accelerations by spinning something round on the end of an arm, but there's always that pesky downward gravity vector that keeps getting added to everything, unless you build your simulator in orbit.

Yes, NASA Ames has had their huge Vertical Motion Simulator for many years. That has about 20 meters of travel in each axis. It used to have a Space Shuttle cockpit mounted; now it's mostly used with a helicopter cockpit. NASA's is a much heavier system, because it has to lug the whole cockpit around.

I agree that rotation should be enough for VR purposes, and linear translation of the subject is not strictly necessary.

Any acceleration up to g = 9.81 m / s^2 can be easily simulated, and it can be maintained indefinitely. All you need to do is to rotate the subject so that the perceived acceleration is aligned with g, or keep it at an angle, to get a fraction of g. Interestingly, for VR you don't need to simulate velocity, as it cannot be perceived by the body.

Flight simulators have been using this approach for decades now, and it works really well: http://arabiansupplychain.com//pictures/300x200/CAE-simulato...

But this system looks to have the advantage of being able to simulate accelerations above g.

PS: Is gimbal a verb now?

Not up to 1g acceleration, exactly 1g aceleration only (if the platform only rotates), with varying direction.

You can't do brief periods of weightlessness with rotations on a gimbal, or periods of larger than 1 g.

Reply on news.ycombinator.com