Check the software usage to get the pose of the trackers without the headsets

[HosameldinMohamed]

Following https://github.com/ami-iit/yarp-openvr-trackers/issues/10, we are able use openVR to obtain the trackers poses without the headset. However, the current version of the software (https://github.com/ami-iit/yarp-openvr-trackers/releases/tag/v0.0.1), doesn't support it yet.

In this issue, we highlight the difference in behavior when eliminating the headset.

In detail, following #10, then launching

yarpdev --device transformServer --ROS::enable_ros_publisher 0 --ROS::enable_ros_subscriber 0

then

yarp-openvr-trackers

Streams these error messages

[ERROR] The state of the output of device LHR-87B324E3 is not ok

Which seems to be coming from https://github.com/ami-iit/yarp-openvr-trackers/blob/cd153dbb2c9a6c5056cf13bd74865c7b14850f1a/src/OpenVRTrackersDriver.cpp#L317-L322

[HosameldinMohamed]

Some observations

• When we launch yarp-openvr-trackers it detects a device called /hmd/Null Serial Number, which seems to be the dummy driver we enable to bypass the headset. It also detects the trackers when they are connected.
• The identified tracker is considered of the unsupported type, seeing this message

[INFO] The device xxxxxx has unsupported type

Which seems to be coming from https://github.com/ami-iit/yarp-openvr-trackers/blob/cd153dbb2c9a6c5056cf13bd74865c7b14850f1a/src/OpenVRTrackersDriver.cpp#L203-L206

But it seems only a warning.

• Commenting out the checks in https://github.com/ami-iit/yarp-openvr-trackers/blob/cd153dbb2c9a6c5056cf13bd74865c7b14850f1a/src/OpenVRTrackersDriver.cpp#L317-L322 and in https://github.com/ami-iit/yarp-openvr-trackers/blob/cd153dbb2c9a6c5056cf13bd74865c7b14850f1a/src/OpenVRTrackersDriver.cpp#L324-L328 doesn't solve the issue. The streamed pose of the tracker seems frozen, probably it took the first value and then refused to take other values.
• Actually, looking at the documentation of the method GetControllerStateWithPose, it seems that the pose is updated when the last button event occurred, so it was probably waiting for a button to be pressed (Maybe it considers the tracker as a controller).
• It seems that the correct method is GetDeviceToAbsoluteTrackingPose, (I didn't see then the issue #8 😄).

[HosameldinMohamed]

• It seems that the correct method is GetDeviceToAbsoluteTrackingPose, (I didn't see then the issue Use GetDeviceToAbsoluteTrackingPose instead of GetControllerStateWithPose #8 smile).

Replacing https://github.com/ami-iit/yarp-openvr-trackers/blob/cd153dbb2c9a6c5056cf13bd74865c7b14850f1a/src/OpenVRTrackersDriver.cpp#L309-L314 With

vr::TrackedDevicePose_t poses[pImpl->devices.size()];
pImpl->vr->GetDeviceToAbsoluteTrackingPose(vr::ETrackingUniverseOrigin(pImpl->origin), 0, &poses[0], pImpl->devices.size());
pose = poses[pImpl->devices[serialNumber].index];

Seems to do the trick! The behavior is as expected. The trackers' poses are streamed via the transform server, and The checks are not throwing warnings.

However the method is a bit different from the previous one because it expects an array of vr::TrackedDevicePose_t.

We need to review it because it throws a segfault when a tracker is switched off for example. There should be a better way to access it!

[HosameldinMohamed]

After the modification in #12, the device can work without the headset, but I had to change the line https://github.com/ami-iit/yarp-openvr-trackers/blob/e6694adeb30d18f279fabd06fd867d76b44b3415/src/OpenVRTrackersDriver.h#L59-L60 From Seated to Standing. Not sure if updating it upstream will affect other use cases.