Long-term running crash

[houlei516]

Dear Jeff,
Thank you for this awesome stack. When we are testing the code, e.g. transfer to 864 candidate views set in both the Kinect version and the Indigo version, with corresponding 864 reconstruction steps, on a specific reconstruction step, the code breaks by 1) failing to receive the pcl data so the demanding data command receives no response. 2) Kinetic: cannot get camera pose when moving to NBV in the 281th iteration. The flying_stereo_cam node breaks up directly.

When we are testing the random sampling which takes several hours to update the free voxel, along with the occupied voxel, this stack cannot run long time.

Thank you.

Both bugs terminate the reconstruction process in the long run in both ROS versions.