Unstable Timestamp Synchronization Between Stereo Cameras in ROS

[sdjtcn]

Hello maintainers,
One:
I'm using this project for stereo SLAM development and encountered an issue with timestamp synchronization between left/right camera images. When acquiring images via ROS1 nodes, the timestamp difference between stereo pairs varies significantly (4–20ms) and fluctuates unpredictably.

Two:
"When acquiring images directly through the Python API, I've encountered two scenarios across multiple script executions:
​​Perfect sync​​: Timestamps from both cameras are identical
​​Significant drift​​: Timestamps show ~50ms discrepancy
This inconsistency persists even when restarting the Python script."

Could you suggest any methods to:
1/Implement software-based synchronization with stable timestamp alignment
2/Recommend best practices for configuring the nodes to minimize this discrepancy
or alternative solutions？

Thank you for your time and expertise!

What feature are you suggesting?

Overview:

Smaller Details:

Nature of Request:

• Addition
• Change
• Removal

Why would this feature be useful?

[jonyMarino]

@sdjtcn This appears to be a bug report rather than a feature request.

Are you aiming to receive high-resolution or large image messages? It’s possible that the NNG socket is being overwhelmed. Which can also cause the Python ROS wrapper to fall behind and drop messages. This is a known limitation.

Addressing this properly will likely require community work on #13, including the development of a more performant C++ ROS wrapper. Contributions and discussion on that front are very welcome.

[sdjtcn]

@jonyMarino Thank you for your prompt reply.

In the scene configuration file, there are a total of four cameras. Among them, the left and right cameras used for SLAM share the same image resolution of ​​640x360 @ (20~40 fps)​​, and the image type is ​​"scene"​​.

robot_quadrotor_px4_sitl_vslam.jsonc