Motion Simulator using the Oculus Rift

[sirnoname]

In simulators you should not play around with recalculating the position in space. It must be mechanically out of the calculation. Else you have delays and calibration problems. DK1 is a nogo.

http://www.heise.de/ct/artikel/VR-Brill ... 50859.html

[jhordies]

Position in space:
We will not calculate the position in space, it will be picked by the camera mounted on the moving rig.
So it will take the position of the user relative to rig not to the ground and this without any customization.

Orientation :
We are recalculating the orientation based on sensor picking the actual orientation of the rig (moving chair).
We don't assume anything we measure the moves of the chair (using the same technique as Oculus) and simply "subtract" them from the ones the Oculus picks.

True, DK1 is not fit for users to play. I think DK2 will be already much better and already usable to actually play.

[xa4123]

hey jhordies,

Very nice !
I've been dreaming of making the same setup ... already have Thanos AMC1280USB board, but haven't made a full scale stewart platform yet and I'm also waiting for my DK2 to be shipped.

Some ideas I wanne share :
You sure the camera in DK2 is only for position ... I was hoping to use the camera for both, position and orientation (ignoring the IMU/MARG sensor data). I believe optical tracking is the way to go for position and orientation on a stewart platform. My first idea was to use a second IMU/MARG sensor and subtract the orientations, but I wonder about the performance and the accuracy of such a setup. I also wonder if you could use the stewart platform's feedback sensor data for that matter instead of a second IMU/MARG.
I'm looking into Valve's solution for position and orientation tracking in their prototype HMD. They use a camera mounted on the HMD and some kind of markers/Qr codes on the walls ... I've seen other similar solutions in the past and hope this will eventually be a cheap and robust solution for optical position and orientation tracking.
My last idea (not reallu my idea ... suggestion someone made on the MTBS forums) is maybe not so elegant, but I think its worth looking into it. Build an enclosed cockpit-like box on top of the stewart platform and use several fine cables connected between potentiometers and the user's head (+ neck + sholders + ...) to determine orientation and position (relative to the cockpit and hence platform) It should also be interesting to use Valve's camera+markers solution in a cockpit.

Good luck with the project.

[tzuvela]

Hi jhordies,

Amazing setup, when it's complete you'll own a closest thing to a holodeck currently in existence!
I am planing to build a very similar 6dof motion simulator for use with DK2, however I'd like to keep the cost low and I am still looking for suitable motors.
These worm drive motors are very interesting:
http://www.motiondynamics.com.au/worm-d ... orque.html
However I am worried they might back drive, so If I go with them I'll have to fit another gearbox, or assemble some sort of lever system to lower workload on the motors.

If I decide that's too much hassle then I'll probably save up for these:
http://www.motiondynamics.com.au/12v-or ... -30nm.html
(I'd probably go for 12v, 64rmp 30Nm torque)

Can you tell us a bit more about the motors you choose, rpm, torque, what were you looking for when you did the research?

Why are you making such a large platform only for a seat?
Perhaps it might be better to make a frame around the seat and attach motors to the frame, something like: https://www.youtube.com/watch?v=zNUBZfrOXUc

Regards,
Tino.

[jhordies]

Hello xa4123,

Thanks for your comments.
Yes I'm sure about the sensors :
http://www.oculusvr.com/dk2/
It says :
Internal Tracking
Sensors Gyroscope, Accelerometer, Magnetometer
Update Rate 1000 Hz

They drastically increased the update rate, I hope our system tested on DK1 will still work smoothly because we don't have such an update rate.
We only managed to have something stable around 100 Hz which was perfect for DK1, we will see what are the results with DK2, we might have to improve it somehow.

Maybe you could guess the orientation with the feedback of the Thanos board but it would require more calculation and they would be quiet complex.
As sirnoname was mentioning, you should avoid calculated values to avoid delays and errors, this is why I wen for a sensor measuring the rig orientation at high speed.
One more thing about using the feedback of the Thanos board, I doubt you can retrieve the values from the board as is. I might be wrong but it wasn't meant for that. I will ask Thanos next time I speak to him.

About the wire solution, wouldn't it be a pain to climb into the simulator and getting hooked to the system ?
I really believe, the Oculus team have covered this part in a better way that what I could mock.

According to the test I did with the DK1 I think it will be just fine using the extra IMU.

I don't know much about the QRcode and camera solution, let me know if you succeed in that direction, I would be really interested.

Thanks again for your comments.

[jhordies]

Hi Tino,

Thanks a lot for the cheering
We also want to keep the costs down, but going for a 6DOF stewart platform implies 6 motors and thats why it becomes costly because its more or less 6 times everything.
I made my design in a way that it could be a 1, 2, 3 or 6 DOF. I will see what I can do once I will have the motors.
About the motors I followed vicpopo sizing, 250W and about 200Nm of torque with a brake resistor.
I asked for several quotes and I finally ordered some drive but I'm not that good in mechanics so I will probably now if I made hte right choice only once built.
I'm taking a leap of faith here

I checked the full motion dynamics project before starting, It looked so promising it's too bad they vanished from the face of the earth once the project finished. I tried to contact them but they never answered.

The difference between the design of FMD is the height of the rig, they managed to lower the center of gravity while I'm trying to use as much out of the shelve products. But I'm quiet pleased of the result, it should take much more space on the ground than the FMD design. I'm limited by the size of the AC motors + gear box.

Best of luck with your project
Regards

[jhordies]

You can now find a video of the offsetting of the Oculus to the chair.
on
http://motiondrift.blogspot.be
or
http://youtu.be/6-WIvSV9438

As you can see the user can still use the Oculus goggles but if the chair moves it is not picked by the game.

[tzuvela]

Hi jhordies, great work or chair offset!

250W and 200Nm wow, that is amazing amount of torque, I'm sure your motors will work great.

I'm trying to find a way to use cheap wiper or worm drive motors.
I'm tinkering with adding a lever between the arm on the motor and pushrod.
Pivot point should help with the weight of the chair and motor would be on the longer side of the pivot point so the workload would be less.

[zerax]

I hope our system tested on DK1 will still work smoothly because we don't have such an update rate.
We only managed to have something stable around 100 Hz which was perfect for DK1, we will see what are the results with DK2, we might have to improve it somehow.

Hi jhordies

Awesome project, have you checked out the trackers from yeitechnology they have much higher update rate.

Edit:

What is the expected mechanical latency from the intermediate program to motion of the platform.

[jhordies]

Hi zerax,

Thanks for the link, it could be a good candidate indeed .
We will first see how good it works with DK2 and focus now on the motion simulator.
But it's definitively on my wish list !

Sorry I didn't understand your question about the expected mechanical latency.