Gazebo#

mocap4r2_gz_plugin (Jazzy and newer) is a world plugin that simulates a motion capture room: it reads the pose of the links you choose, adds markers, and streams the frames

  • as a MOCAP4ROS2 system (output: ros), with simulation time, or

  • as OptiTrack Motive (output: natnet): the real OptiTrack driver connects to Gazebo exactly as it would to Motive. Useful to test a whole setup, including the driver and its network configuration, before going to the lab.

$ ros2 launch mocap4r2_gz_plugin mocap_room.launch.yaml

Configuration (in the world SDF):

<plugin filename="mocap4r2_gz_plugin" name="mocap4r2_gz_plugin::MocapSystem">
  <output>ros</output>          <!-- or natnet -->
  <rate>120</rate>
  <noise>0.0003</noise>         <!-- marker noise std, metres -->
  <dropout>0.0</dropout>        <!-- probability of losing a body in a frame -->
  <autostart>activate</autostart>
  <frame_id>mocap</frame_id>
  <rigid_body name="robot" link="my_robot::base_link">
    <offset>0 0 0.3 0 0 0</offset>   <!-- body frame in the link frame -->
    <marker name="front">0.12 0 0.02</marker>
    <marker name="left">-0.06 0.08 0.02</marker>
    <marker name="right">-0.06 -0.06 0.04</marker>
  </rigid_body>
</plugin>

With output: natnet, natnet_address, natnet_command_port, natnet_data_port, natnet_multicast and up_axis (y like Motive by default) configure the simulated Motive.

The plugin does not publish while the simulation is paused, restarts its frame numbers when the simulation is reset, and stamps the data with the simulation time.

Demos#

mocap4r2_gz_demos has two ready-made scenarios:

$ ros2 launch mocap4r2_gz_demos robot_groundtruth.launch.yaml system:=ros
$ ros2 launch mocap4r2_gz_demos robot_groundtruth.launch.yaml system:=optitrack
$ ros2 launch mocap4r2_gz_demos human_motion.launch.yaml

  • Robot ground truth: a robot with a marker plate drives in circles; robot_gt publishes base_link_gt. With system:=optitrack Gazebo plays Motive and the real OptiTrack driver connects to it, as in the lab.

  • Human motion: a walking mannequin streamed as a skeleton (and its head as a rigid body); skeleton_publisher publishes its segments as TF and its joint angles.