VIVE trackers and arthroscopic surgery simulator

[papadonik]

We are a startup company and have managed to build a VR surgical training system that tracks motion accurately and without "tremor" using tracking algorithms for the area of interest which is about 50 x 50 x 50 cm. The motion is excellent, the Vive newest trackers are attached on real surgical instruments and the user has haptic feedback from the surgical instruments along with the VR experience. These instruments go into a plastic anatomic model - human body- through holes. There is a snapping process on the software side that allow us to have a fixed point of rotation like in real endoscopic surgery. MOST OF THE TIME the system works fine. Occasionally, when for example the computer goes into sleep mode, or we restart the computer, this excellent accuracy which is within mm is lost and the tracking is off by 3 cm. This happens consistently only on side of the plastic model, let's say the right side. Our question : how can this be random? Is there a way to avoid it? We have invested 2.5 years in R&D and everything else works perfectly. Could it be related to the Accelerometer's g range (embedded into Vive) ? I would assume that for the slow movements we are doing a low g range is required? Any input will be appreciated?

 

[JoeJ]

Just out of curiosity, are you using lighthouse based tracking for this app?  Seems like you would have to dial in a very precise pose in the universe for those lighthouses in order to get everything that accurate.

[papadonik]

Yes, we use 4 base stations. How do you "dial in a very precise pose in the universe"?

[JoeJ]

The app sounds pretty awesome.  Lighthouse based tracking is quite accurate but the lighthouse driver can be a bit of a mystery at times.

By dialing in the pose, I just meant that you can get slightly different poses for your root lighthouse during startup of the environment.  If you watch the universe poses in lighthousedb.json between runs you may notice that the target pose from the root lighthouse at (0,0,0) to your secondary lighthouses may be slightly altered during startup or even after running for a period of time.   Getting precise positioning highly depends on how well your tracker is seeing the lighthouses during startup, if there is any reflection in the enviornment, etc.  You might want to copy off that file from one run to the next and compare them to see if there is any significant movement in the universe poses.  Also, I have noticed that the initial pose that lighthouse driver calculates during startup can makes a difference on how accurate your position is.  It would probably depend on whether your device is stationary, the calculated pitch/roll of the universe, etc.   Theoretically the kalman should dial in that pose after a period of time but I've noticed that sometimes an initial pose or pitch/roll that is slightly off may cause a visual difference (i would check the logs between runs to see if there are any explainable differences).

 

[papadonik]

I will look into it. thank you for the explanation. It is appreciated.