FAQ
Content:
Some of our current dependencies include further catkin packages, which are not a dependency of maplab. If you build the complete workspace, these packages will fail due to missing dependencies of their own. This might change in the future, but for now there are two options:
Please only build maplab: catkin build maplab
or
Add CATKIN_IGNORE packages to the superfluous packages.
The decision to move the dependencies to a separate repository was motivated by the fact that someone might want to integrate maplab in their existing catkin workspace. This will require some additional effort, however, having an external dependencies-repository should make it easier to combine the maplab dependencies with the dependencies of the existing workspace.
If you get an issue similar to this:
make[2]: /home/user/maplab_ws/devel/.private/maplab_common/bin/protoc: Command not found
make[2]: *** [compiled_proto/maplab-common/id.pb.cc] Error 127
make[2]: *** Waiting for unfinished jobs....
Make sure that your workspace layout is merge:
cd ~/$CATKIN_WS
catkin clean # Necessary to clean up your workspace as your layout will change.
catkin config --merge-develYou can check if your workspace is properly set up by running:
$ catkin config
---------------------------------------------------------------------------
Profile: default
Extending: [explicit] /opt/ros/kinetic
Workspace: /home/user/maplab_ws
---------------------------------------------------------------------------
...
---------------------------------------------------------------------------
Devel Space Layout: merged # <-- This should say `merged`.
Install Space Layout: None
---------------------------------------------------------------------------
...A known cause for this is an installed MATLAB using a network licence. The build can stall on getting the licence authenticated; so make sure that the licence server can be accessed, connect to the VPN if required.
make[5]: *** No rule to make target '/usr/lib/x86_64-linux-gnu/libGL.so', needed by 'lib/libopencv_viz.so.3.2.0'. Stop.
consider redefining the symbolic link that somehow got lost in your Ubuntu by:
sudo rm /usr/lib/x86_64-linux-gnu/libGL.so
sudo ln -s /usr/lib/libGL.so.1 /usr/lib/x86_64-linux-gnu/libGL.so
Install requests by calling:
sudo pip install requests
Q: Why are even some system dependencies wrapped into catkin packages? (eigen, protobuf, glog, opencv, ...)
Since we support both Ubuntu 14.04 and 16.04 and even have some OSX users (no official support), the library versions provided by the system are varying too much and made developing compatible code a very tedious business. That's why we wrapped some of the system dependencies into catkin packages to gain full control over these libraries.
Most catkinized dependencies will first download an archive or checkout a git repository. If your internet connection is unstable or very slow these dependencies might fail to build.
If you are unable to get access to a stable internet connection, you can try to manually download the zip files/checkout the git repositories and copy them in the appropriate folder. You will find the download/git link and the target folder in the CMakeLists.txt file in the ExternalProject_Add block.
Example - opencv3_catkin:
SET(OPENCV_SRC_PATH "opencv3_src")
ExternalProject_Add(opencv3_src
DEPENDS ${CONTRIB_NAME}
URL https://github.com/Itseez/opencv/archive/3.2.0.zip # <========= LINK TO ARCHIVE
URL_MD5 bfc6a261eb069b709bcfe7e363ef5899
UPDATE_COMMAND ""
SOURCE_DIR ${OPENCV_SRC_PATH} # <========= TARGET FOLDER
BINARY_DIR opencv3_build
PATCH_COMMAND patch -p0 < ${CMAKE_SOURCE_DIR}/fix_python_discovery.patch
CONFIGURE_COMMAND cd ../${OPENCV_SRC_PATH} && cmake- Download the archive from:
https://github.com/Itseez/opencv/archive/3.2.0.zip - Extract the opencv folder to:
~/maplab_ws/build/opencv3_catkin/opencv3_src
It seems that you have checked out an incorrect commit of Voxblox. Maplab requires a specific commit that guarantees compatibility.
Please make sure you have followed the installation instructions, specifically the line:
https://github.com/ethz-asl/maplab/wiki/Installation-Ubuntu
This should clone the correct versions of all the dependencies.
Alternatively, you can go to maplab dependencies repository:
https://github.com/ethz-asl/maplab_dependencies
and make sure your local commits of dependencies adhere to the ones in master branch.
We are supporting the Skybotix VI-Sensor, Google Tango tablets and the Intel Realsense ZR300 by default and provide example calibrations for them here and here.
Maplab focuses on visual-inertial mapping, therefore, the minimum requirements are: a gray-scale (preferably fish-eye) camera with global shutter and an IMU with similar or better characteristics as the ones in the sensors above (e.g. ADIS16488 or BMX055). The second critical component is a solid time synchronization between camera and IMU, i.e. ideally both devices are triggered based on the same clock. And finally, you will need to properly calibrate your sensor as described here and here.
Please also have a look at the answer to the issue(s) below. Related issues: #19
Use the Tango ROS Streamer app to publish the sensor data to ROS and record a rosbag.
We've tested this sensor and once properly calibrated, it works well with maplab. We provide a simple ROS node to use the sensor in this repository (maplab_realsense). For an example calibration and more details, see the ZR300 wiki page. Related issues: #32
Use the --vio_camera_topic_suffix flag to change or remove the suffix when running ROVIOLI:
--vio_camera_topic_suffix="" # To remove the suffix completelyUse the --rovio_enable_frame_visualization=false flag to disable the output.
Most of the commands (such as optimization, loop-closure, visualization, etc.) only work on landmarks that are flagged as 'good'. The default parameters for these quality metrics might not be optimal for your use-case. Have a look at the following page and check if some parameter need to be adapted for your use-case: https://github.com/ethz-asl/maplab/wiki/Optimizing-VI-Maps
That's why:
