How do you use ROS bag files for robot debugging?

When doing remote deployments of a huge number of robots, some robots might behave differently and the exact reason why that is can be told using ROSbags. A lot of new developments and improvements can be done as these bags show almost every valuable data of the robot motion. Additionally, a lot many times mechanical issues can also be estimated from ROSbags. One such example I can give is, in one of our projects we had a robot going towards the left always and we never understood why. When we looked at the bag one of the motors was consuming more power than other, thus leading us to the doubt that it was something to do with Gearbox and it was indeed a Gearbox issue. ROS bags are a huge gift to Robotics Developers and Site Engineers.

ROSbag files allows developer to easily develop algorithms offline. If working on SLAM problem, record a ROSbag of all the sensor observations, odometry values and this could be used for creating a map offline and optimize the mapping tool. This could also be a source of training data if you are using Computer Vision models for obstacle detection or human pose detection.

Recording ROS bags is a very powerful way that allows us to capture the state of things such as sensor readings, robot state, and the robot's surrounding conditions in a given period in time by subscribing to one or more ROS topics and storing the received message data in an efficient file structure. Apart from being used for performing analytics on these topics or debugging your robot system, these can be used as input to unit test your software and in turn enhance the development cycle by not relying on simulation data but using real world data to directly test future releases of software.

ROS bag stores data streams from the robot's sensors into binary file/s. You may also customize the streams to limited sensors or sources in general. The records simulate real-time data. For the robot, that allows you to debug/test/upgrade certain features and re-visualize the output without involving the robot.

ROS bags are life saver when your application is rare to test upon. While we were doing sanding on planes, the planes were coming in hanger once a while which lead us to prepare testing of new things and old things packed in only one firmware. At that momennt we couldn't fix the problem while the application was going on. Fortunately, we recorded data from ros nodes in ros bags and played afterwards. We made fixes and new features and tested on data of ros bags which helped development process. Git was one of the helpful things which helped us keeping master and new features apart for development.

What should be considered before recording a Ros bag? If you want to debug or record a Ros bag for testing specific features, you should verify that the target sensors don't face any noise beforehand. Preparation (Sensor noise validation): You should select desired topics to record and determine which sensors you are considering. Then, you'll need to run the robot on the intended environment and echo the chosen topics to make sure that it gives you valid readings. Some sensors can have huge noise based on spatial location, like GPS. If you start recording a Ros bag without verifying the sensor data, you will repeat the entire process! Be careful here. Running a robot and collecting data, particularly a huge one, isn't an easy task.

Tutorials often mention how to record every message published in your system but most of the time you only need specific ones. You can use the command without the -a flag to record specific topics like this: rosbag record -O bagname.bag topic1 topic2 "/ns1/topic1" You can also use regex to match which topics you want to record. For example: rosbag record -e "/(.*)_stereo/(left|right)/image_rect_color"

Recording ROS bag files is straightforward. Start by choosing which topics you want to capture and where to save the file. For example, to record all topics and save them in a file named my_bagfile.bag, use: bash rosbag record -a -O my_bagfile.bag This command grabs messages from every topic and stores them in the specified file. Press Ctrl+C to stop recording when you're done. If you want to record only specific topics, list them after the record command, or use wildcards and regular expressions for more control.

There are different ways to record ROS bags. One way, as mentioned, is to use a command line terminal to record full streams or specific topics (streams). However, another way is to do it through a package (node) by adding `<node name="rosbag_record" pkg="rosbag" type="record" args="--all --output-name=<PATH_TO_SAVE_THE_BAG>" />` in the launch file. This approach facilitates/organizes providing a customized recording for the bags. In massive projects, recording a rosbag through a terminal might be challenging, especially in the case of customized recording (Not recording everything). Remember that recording sensor data requires a huge store, normally, we record ONLY the desired sensor based on the purpose.

Even though using the "play" terminal command is a perfectly fine method of playing ROS bags, I prefer to use a third-party tool called plotjuggler. It allows me to record and replay ROS bag files and convert them into other forms of data. It also allows plotting them, editing and processing them live.

To play ROS bag files, use the rosbag play command followed by the file name. For instance, to play my_bagfile.bag, run: bash rosbag play my_bagfile.bag This command replays the recorded messages as though they’re coming from the original sources. You can adjust playback with options such as: --rate <factor> to change playback speed --loop to replay the file continuously --pause to start playback in paused mode --start <time> and --duration <time> to skip to and play specific sections These options help tailor the playback to your needs for debugging or analysis.

One of the easy way is extract the rosbag files and use the pkg rosbag to csv You will have csv files and you can see the particular timestamp as well as the events occured at that time stamp Soo if you know the error or the time the error occured you can get the csv files and understand the behavior of that time

When working with ROS bag files, it's not just about isolating specific topics; sometimes, you might want to filter messages based on their content or timestamp. For instance, if you're troubleshooting a robot's behavior and you know the exact time when an anomaly occurred, you can filter messages around that time frame. To filter messages based on their timestamp, you can use the --start and --end options followed by the desired time. Furthermore, if you want to filter based on the content of the messages, you can use Python expressions. For example, if you have a topic /robot/position that publishes geometry_msgs/Point messages and you can filter out messages where the x value is greater than 10.

To filter ROS bag files, use the rosbag filter command to extract specific messages based on your criteria. This helps in isolating data from certain topics or time periods. For instance, if you want to extract messages only from the /camera/image_raw topic, run: bash rosbag filter my_bagfile.bag camera_only.bag "topic == '/camera/image_raw'" This command creates a new bag file named camera_only.bag containing only the messages from the specified topic. You can also use more complex Python expressions to filter messages based on time or other conditions, making it easier to focus on the data that matters most for your analysis.

Two notes on issues when working with bag files (in ROS1 (kinetic, Melodic, Noetic, etc) Be sure to check whether or not to use simulation time (typically a rosparam, use_sim_time) in your application. This is especially important when working with multiple bag files simultaneously as it's essential to ensure that captured data is processed in the intended order. Each bag file records the time of messages and can either replay them as a live stream (same offset from the time the bag is captured to the time it is played), or data collected at that particular time. Namespaces of all messages/topics captured by the bag are preserved so remapping may be needed to tie into other packages (especially if the bag is being used as a unit test)

To dive into your ROS bag files, you have a few handy tools: 1, rqt_bag: This is like your bag file’s personal assistant. It lets you see and plot messages in a user-friendly graphical interface, making it easy to visualize and edit data. 2, rosbag info: Think of it as the bag file’s resume. It gives you a quick overview of essential details like size, duration, and the topics covered. 3, rostopic echo: This tool lets you peek inside specific topics and see the messages, with options to control how much or how often you view them. 4, rosbag API: For those who like coding, this Python or C++ API allows you to create custom scripts to analyze or manipulate your bag file data.

Found rosbags are great for debugging and analysis purpose, but personally i feel it occupies large amount of memory while storing

Points to consider when recording a ROS bag: Sensor Configuration: Ensure that all sensors are correctly configured. For instance, if a camera's exposure settings are off, the recorded images might be too dark or washed out. Timestamp Synchronization: If you're using multiple sensors, ensure that their timestamps are synchronized. This is crucial for tasks like sensor fusion or when analyzing the interplay between different data sources. Topic Selection: Instead of recording all topics, which can quickly fill up storage, select only the topics relevant to your analysis or task. This not only saves space but also makes post-processing more manageable.

Everyone has provided a very comprehensive and informative overview of how to use ROS bag files for debugging. I have a few additional tips that you may want to consider: Use descriptive names for your ROS bag files. This will make it easier to find the right bag file when you need it. For example, instead of naming a bag file bag1.bag", you could name it "robot_navigation_test_2023-10-12.bag". There are a number of ROS bag management tools available, such as rqt_bag and bagman. Use a ROS bag recording framework. A ROS bag recording framework can help you to automate the recording and processing of ROS bag files. This can be useful if you need to record a large number of bag files or if you need to record bag files in a specific way.

What should be considered while recording a Ros bag? Although validating sensor noise beforehand is a crucial step that should be done before recording a Ros bag, as I explained earlier, you still need to be careful while recording a Ros bag. Otherwise, you may repeat the entire process! One example illustrates this case: Assume you are recording a Ros bag for mapping a certain region/place using a LiDAR sensor. If the region is highly dynamic, say having many people moving around, you wouldn't be able to use the recorded point cloud to generate a clear map as it will have a lot of noise. Also, the noise might have come from your reflection on the point cloud! The bottom line is that having a proper plan before starting.

Be careful of recording video data. ROS publishes video streams in individual frames, most of the time uncompressed. This means you will record every pixel of every frame which will fill a whole hard drive pretty quickly!