ROS master unexpectedly kills without error (when using custom launch file)
Hi all,
Config: I am dealing with a Husky A200 robot from clearpath robotics, running Ubuntu 18.04 and ROS Melodic. The sensors (which I am using) are a ZED stereo camera, a FLIR monocular camera and an IMU.
Goal: Simultaneously record the left and right images from ZED and some diagnostics topics into multiple rosbags. I'm trying to achieve this using a launch file I wrote myself. I'll link it here.
<!-- -*- xml -*- -->
<launch>
<arg name="record_zed_rgb" default="false" />
<arg name="seqno" />
<arg name="output_dir" default="/media/5944fb55-8975-45e7-aaff-436bc493fe61/results/" />
<node pkg="rosbag" type="record" name="rosbag_record_zed_rgb_rect"
args="record -O $(arg output_dir)seq$(arg seqno)_zed_rgb_rect_images /zed_node/rgb/image_rect_color"
if="$(arg record_zed_rgb)" />
<node pkg="rosbag" type="record" name="rosbag_record_zed_params"
args="record -O $(arg output_dir)seq$(arg seqno)_zed_params /zed_node/depth/camera_info" />
</launch>
I then run: roslaunch <launchfile_name.launch> seqno:=000 record_zed_rgb:=true
The problem doesn't occur when I omit the following from the second rosbag record instance.
/zed_node/depth/camera_info
Do you guys have any idea why this happens? I can visualise a depth point cloud through rviz, so it's not the depth node that's broken.
I have run into random kills at other times, for example when using the rqt's visualisation tool and subscribing to more than 1 image view. Or when trying to move the robot (joy_teleop using PS4 controller) while recording images.
Thanks in advance!!
Could you share
rosbag info <your bagfile>
Sorry for the delay, it was in the middle of the night when I read the notification and I had forgotten about it.
I just recorded this:
What are your hardware specs? Based the manual, the computer for A200 robot is from 2011-2012 and I see many people are upgrading Jetson Nanos. Based on your
rosbag
you are processing images and maybe running out of memory if each msg is in the 3.5MB size. One approach might be to work with compressed images, or reduced the frequency of sampling.Dear Osilva, thanks for taking the time! I'll have a look when I have access to the robot tomorrow. I know that it's an Nvidia Xavier Jetson, but I don't know which specific model and/or what specs.
I can also try to verify by recording to internal memory and see how it responds. Now, I've been trying to record to a USB thumbdrive, which could also easily be a limiting factor. It doesn't complain about a overloading buffer, however.