Vive tracker vibration update?

[Jmels]

Curious to know if there is any update to correct imu data for Vive trackers (2018) so prolonged vibrations do not skew tracking data. I know the 2017 tracker had the low pass filter to correct this, but is there any plan to add this to the current tracker? Without proper tracking, we'll probably have to end up abandoning the Vive trackers. 

[Rich_BlackBox_VR]

+1

[HackPerception]

@Jmels @Rich_BlackBox_VR,

No - LPS support is not road-mapped for Vive Tracker 2019 due to changes in how event and haptics triggers work within SteamVR input 2.0 and the underlying SteamVR hardware firmware architecture. The old LPS methodology is no longer valid and breaks tracking under the new system. The new system apparently means that if we were to implement LPS, it would be very difficult to implement on the developer side to the point where it'd be impractical since LPS requires the event trigger to be sent to the tracker prior to a haptic event.

There's a LBE company that's come up with a physical isolation methodology we're recommending for professional use cases. I'm checking now to see if I am allowed to refer you to them or if it's behind NDA.

[RobTaylor@PI]

On 2/28/2020 at 4:10 PM, VibrantNebula said:

@Jmels @Rich_BlackBox_VR,

There's a LBE company that's come up with a physical isolation methodology we're recommending for professional use cases. I'm checking now to see if I am allowed to refer you to them or if it's behind NDA.

VibrantNebula -  Were you unable to find out about this LBE company? We're in development on a high-profile VR experience using haptic guns which cause significant tracking issues with our Lighthouse 2.0 setup using Tracker (2018). We're desperately seeking a solution or will be forced to abandon Lighthouse tracking and Vive tech. If you're not able to post this company's name publicly, would they be willing to allow you to connect us if we sign an NDA? ...or would they be willing to hear about our needs through your team, and decide if they'd like to contact us? We are happy to pay for an effective solution.

Regards,

Rob

[JMS3DPrinting]

Is there any upgrade to the Vive accelerometer to allow higher vibrations without losing its mind and floating off?

It appears to do this at anything higher than 8 g's.  That is extremely low limit that is easily exceeded just by bumping it.

That should be an achievable upgrade without having to change code.  I would have hoped the 3.0 version would have included a more robust accelerometer.

[HackPerception]

@JMS3DPrinting SteamVR has a specific relationship with the IMU because the basestations can only output data at ~120hz interleaved (behavior is more complex with 2.0 stations). The IMU data is required for sensor fusion because the base stations provide updates at ~1/10 of the speed the IMU can so the IMU plays a huge role in micro-adjustments and interpolating for when there aren't enough optical samples for a pose. A strong g-shock really confuses the sensor fusion and it can take a few frames of optical data for the tracking system's to get enough optical samples to create a pose estimate that makes sense for both what the optical sensors are seeing and what the IMU is reporting.

It isn't that this is a Vive-specific thing. It's more of a low level SteamVR tracking thing and past a certain depth, SteamVR tracking is a blackbox to everybody outside of Valve. You'd hit a similar issue with other SteamVR hardware, but the specifics will probably vary a little. Similarly, optical tracking systems have similar limitations (with the additional limitations that the tracking cameras operate at less than 120hz). Every tracking system will have a different way to try and smooth out all of the uncertainty from when the the tracking system updates slower than the device is moving through physical space. SteamVR uses less of the predictive modeling that Oculus - alot of Oculus tracking is filtered through machine learning and isn't representative of the real sensor values.

When dealing with high-G force or high-speed tracking scenarios, using a high-speed imaging system like Optitrack may be the only option in many high-performance use cases.

@JMS3DPrinting It's not a case of the IMU not being robust - it's a case of the other tracking components (basestations) working at a fraction of the refresh speed of the IMU. The IMU can output data that's has such a high delta that the sensor fusion can create a valid pose estimate and it needs to hit enough optical samples that it can properly place itself again.

First generation Vive trackers had a low pass filter firmware option that could provide some level of buffer to help smooth out G-spikes but that same approach is not possible with newer SteamVR tracking implementations due to changes in the SteamVR hardware stack.

If you have a situation where you're regularly getting over 8g or move the controllers faster than SteamVR can track - you're really in the realm of Optitrack and other high-demand solutions.