[GSoC] Project #9 discussion: ROSbag Reader Node — dataset and message scope

[DGHX12345]

Hello dora-rs team,

I am very interested in Project #9: ROSbag Reader Node for Dora for GSoC.

I am currently involved in an open-source robotics project and have worked with camera data and robot-arm joint-related data, so I am especially interested in improving offline data replay in Dora.

My current understanding is that this project should let users replay ROSbag data into Dora without depending on a full ROS runtime, so that recorded datasets can be reused more easily for testing, demos, and offline pipeline development.

To keep the first version realistic, I am currently thinking about a narrow MVP that focuses on:

local playback on Linux

a small set of useful message types

one reproducible example pipeline

I wanted to ask a few scope-related questions before I finalize my proposal:

1.For the MVP, which message types would be most valuable to support first?

2.Would it make sense to prioritize camera/image data and joint-state data as an initial subset?

3.Is there a preferred ROSbag format or dataset style that the project should target first?

4.Would the community prefer a smaller first version with fewer dependencies, even if message coverage is limited?

My current preference is to make the first version small and reliable, then leave broader format/message support for future extensions.

Thank you!

[RuPingCen]

Thank you for your interest.

Data from cameras, LiDAR (point cloud), IMU, and GNSS are all required. This tool aims to support reading both open-source and private datasets (rosbag). You can use open source autonomous driving data sets for testing, such as KITTI TUM EUROC.

[1] KITTI: http://www.cvlibs.net/datasets/kitti
[2]TUM:https://vision.in.tum.de/data/datasets/rgbd-dataset/download
[3]EUROC: https://projects.asl.ethz.ch/datasets/doku.php?id=kmavvisualinertialdatasets#downloads

[KOKOSde]

For the first version, I would keep the scope very small. Support one bag format, local playback, and only the message types that unlock a real demo. sensor_msgs/Image, sensor_msgs/CameraInfo, and sensor_msgs/Imu are the best starting point. If you want one more, add nav_msgs/Odometry. That already covers a useful camera plus IMU replay pipeline without turning the project into a general ROS compatibility layer on day one.

[DGHX12345]

Thank you both for the detailed guidance — this is very helpful.

It makes sense to keep the first version focused and avoid turning this into a general ROS compatibility layer. I really like the idea of starting from a small but useful subset that can support a concrete demo.

Based on your suggestions, I will narrow the MVP to:

local playback of a single ROSbag format

a minimal set of message types, prioritizing:

sensor_msgs/Imag

sensor_msgs/CameraInfo

sensor_msgs/Imu

(optionally)nav_msgs/Odometryif time allows

This aligns well with a practical camera + IMU replay pipeline, which I think is a strong and realistic starting point.

For datasets, I will take a look at KITTI, TUM, and EUROC and try to base the example pipeline on one of them so that the demo is reproducible.

I am currently thinking about the MVP workflow as:

1.load a local ROSbag file

2.select relevant topics

3.replay them into Dora

4.provide basic control such as start/stop (and pause/resume if feasible)

5.connect the replayed data to a simple downstream Dora pipeline

I have a couple of follow-up questions to make sure I stay aligned with expectations:

For the first version, would you prefer prioritizing clean playback behavior and integration with Dora pipelines, even if message coverage stays limited?

For the example pipeline, is it better to focus mainly on demonstrating sensor replay itself, or to include a slightly richer downstream use case (e.g., a simple processing node)?

For playback configuration (topics, rate, etc.), is there a preferred way to expose this in Dora (e.g., node config vs. dataflow-level config)?

I will refine the proposal based on this direction. Thanks again for the clear suggestions!