- Feb 19, 2025
-
-
Jan Schmidt authored
Adjust LED model and reported poses to/from IMU-relative as needed (even though the difference is small) Adjust the grip pose back by 10cm
-
Jan Schmidt authored
Fix use of os_mutex_unlock() instead of os_mutex_lock() in a few places. Update the ctrl->pose position when optical updates arrive, so the debug UI shows the position changing
-
Jan Schmidt authored
Pass camera poses and controller models in OpenCV coords. Later this should be reverse, and WMR should flip models and cameras, so the constellation tracker doesn't have to
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
Controllers are tracked using the same origin as the headset now
-
Jan Schmidt authored
Rearrange the camera order so front cameras are first in the lineup, to make it easier to debug things
-
Jan Schmidt authored
Allow the constellation tracker to retrieve the generic tracker pose for it to match LEDs against
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
Not yet connected to the controllers themselves
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
Remove all dependency on the WMR driver from the constellation tracker
-
Jan Schmidt authored
Constellation tracking only notifies about the frame start times now. It's up to the controllers to calculate timesync from that
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
Improve rejection of spurious blobs from windows and house lights by doing a simple size check on the approximate size we expect controller LEDs to appear
-
Jan Schmidt authored
Add handling of WMR controllers and index emulation entries in the SteamVR plugin
-
Jan Schmidt authored
Only apply part of a yaw update for larger updates, to nudge the controller in the right direction and filter out glitches.
-
Jan Schmidt authored
If a controller is idle long enough or restarted, then the device time can change enough to cause the monotonic time matching to go backward, and lead an assertion in the 3dof fusion. Catch the case and reset the clock estimation and fusion entirely.
-
Jan Schmidt authored
-
Jan Schmidt authored
Switch the camera poses, but not the ROI
-
Jan Schmidt authored
Tilt controller grip pose 40° and move it back 10cm
-
Jan Schmidt authored
Add a refinement step, but disable - the OpenCV Ransac PnP frequently produces really wrong answers for some reason.
-
Jan Schmidt authored
Allow for slightly less rigid matches
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
Until a space overseer is set up, override the tracking origin for all the devices the builder creates
-
Jan Schmidt authored
Correct WMR controller tracking transforms by doing all tracking work in OpenCV coordinates, with XYZ Right-Down-Away, 180° flipped from OpenXR/OpenGL coordinates. The controller LED models are already in the right pose, as are WMR configuration poses for the cameras, so only the OpenXR source poses need to be flipped before use and flipped back before publishing. Also improve the debug render and other improvements like not projecting LEDs into a view if they're outside the distortion projection radius or behind the camera.
-
Jan Schmidt authored
But let the gravity swing be determined solely by IMU for now
-
Jan Schmidt authored
Output the latest pose for debug UI like the HMD does
-
Jan Schmidt authored
-
Jan Schmidt authored
Ignore the swing vector and only conside the swing angle for gravity alignment checks
-
Jan Schmidt authored
-
Jan Schmidt authored
-
Jan Schmidt authored
-