I tried to use simulated PR2 and Object Recognition Kitchen. There is some problem with encoding of depth map (topic /head_mount_kinect/depth/image_raw). Not sure if it's problem of the PR2 simulator or ORK (or Gazebo?) so I'm also reporting it here. ORK issue: wg-perception/object_recognition_core#35. I worked with depth maps in ROS/Gazebo long time ago and I remember there were some difficulties with encodings.