aharmat's profile - activity

Please see updated question. Users might need to change launch files in order to tweak parameters or to remap topics. The "user files" in this case refer to camera masks used to block image regions corresponding to the robot body, which will be unique for each user/robot.

I am converting a package from rosbuild to catkin, which has a 'launch' folder as well as folders for various user generated files that are loaded by an executable after the file names are passed in as params. After catkin builds the package, the results are placed in the devel folder, but the launch files and the user generated files remain in the source folder. Using my package in this way seems wrong, as I am constantly messing around with the launch files and the user files, which seems to go against the spirit of the catkin build system. What is the accepted way of handling launch and user generated files with catkinized packages?

UPDATE: I have followed the method at the bottom of http://wiki.ros.org/catkin/CMakeLists... and can now install launch and other files to CATKIN_PACKAGE_SHARE_DESTINATION when invoking the install target on catkin_make. The problem now is that calling roslaunch my_package my_launch.launch still refers to the launch file in the src directory, so the installation was effectively useless. Also, calling $(find my_package) in a launch file finds my_package in the src directory, so any user generated files placed in install/share/my_package are not found.

UPDATE 2: I forgot to source the setup.bash file from the install directory (instead of the devel directory), which is why I was seeing the behaviour above. Once I sourced the proper setup.bash file, everything worked as intended. On a side note, why is CATKIN_PACKAGE_BIN_DESTINATION lib/my_package rather than bin/my_package?

Thanks Markus, this looks like what I need. I'll try to write a node relays all topics from a given machine, essentially acting as a proxy for the aircraft.

In our current setup, we have a laptop connected wirelessly to our robot, everything running ROS. The laptop does most of the processing and sends commands back to the robot, pretty standard stuff.

However, we have noticed that if we are recording all (or even some) of the data being sent by the robot (with rosbag record) while also doing the usual processing, the laptop slows down intermittently, negatively affecting the rate of control commands being sent out.

As a potential solution, we would like to use two laptops, one for data processing and control command generation, and the other one for data recording and maybe data visualization. This would also nicely split tasks between people, with the controller laptop operator focusing on running his algorithm an the data recording laptop operator making sure the experiments are running as designed.

However, I am concerned about the data being sent wirelessly being needlessly duplicated if it has to go to both laptops. The setup I envision has the two laptops connected (wired) to a network switch, along with our wireless radio (a Ubiquity Networks NanoStation) so everything is on the same subnet. On the other side of the wireless connection we have a Ubiquity Networks Rocket connected to a single PC also on their own subnet.

Is my concern warranted, will the data be sent twice if both laptops are subscribing to a topic? Some of the data are images from various cameras so double sending this would be very bad. Is there any way to resolve this problem other than having one of the laptops relay messages to the other (so the second laptop would only subscribe to relayed topics on the first laptop)?