Thank you for providing such a sophisticated open-source framework for multi-sensor fusion.
I am currently working on a project using FAST-LIVO2 and have encountered some technical challenges regarding sensor synchronization and the resulting scan accuracy.
I would appreciate your expert guidance on the following points:
- Hardware Synchronization Setup
To achieve precise triggering, I implemented a hardware sync module using a Nucleo F103RB board with HAL Drivers, referencing the logic from the LIV_handheld repository.
This setup is intended to synchronize the Camera, IMU, and LiDAR.
• Question:
Is this hardware-level approach (F103RB/HAL) generally sufficient for the timing requirements of FAST-LIVO2, or are there specific jitter/latency thresholds I should aim for?
- Analysis of Time-Lag and Sync Warnings
During the operation, I observed a "Sync warning" between the IMU and LiDAR with a specific pattern:
• During Bag Playback: The time-lag between sensors consistently increases.
• Post-Playback Processing: Once the bag file ends, as the system processes the remaining data in the buffer, the time-lag gradually decreases.
• Result: Despite these warnings, the system still manages to generate scans (as shown in Figures 3 and 4).
3. Scan Accuracy and Calibration Necessity
Given the symptoms mentioned above, I am concerned about the reliability of the final point cloud.
• Primary Question: In your experience, can the "increasing time-lag" during playback significantly degrade the spatial accuracy of the scan even if the system doesn't crash?
• Calibration Inquiry: Should I prioritize re-performing the Intrinsic and Extrinsic calibration for all sensors to resolve these accuracy/sync issues? Or is it more likely that the spatial misalignment I see is a direct byproduct of the temporal (timestamp) drift rather than poor calibration parameters?
I have attached screenshots of the generated scans (Photos 3 & 4) for your reference. I would be very grateful for any insights you could provide on whether my focus should be on refining the hardware sync code or on the calibration process.
Thank you for your time and for your contribution to the robotics community.
Best regards,
HJ Kim
Thank you for providing such a sophisticated open-source framework for multi-sensor fusion.
I am currently working on a project using FAST-LIVO2 and have encountered some technical challenges regarding sensor synchronization and the resulting scan accuracy.
I would appreciate your expert guidance on the following points:
To achieve precise triggering, I implemented a hardware sync module using a Nucleo F103RB board with HAL Drivers, referencing the logic from the LIV_handheld repository.
This setup is intended to synchronize the Camera, IMU, and LiDAR.
• Question:
Is this hardware-level approach (F103RB/HAL) generally sufficient for the timing requirements of FAST-LIVO2, or are there specific jitter/latency thresholds I should aim for?
During the operation, I observed a "Sync warning" between the IMU and LiDAR with a specific pattern:
• During Bag Playback: The time-lag between sensors consistently increases.
• Post-Playback Processing: Once the bag file ends, as the system processes the remaining data in the buffer, the time-lag gradually decreases.
• Result: Despite these warnings, the system still manages to generate scans (as shown in Figures 3 and 4).
3. Scan Accuracy and Calibration Necessity
Given the symptoms mentioned above, I am concerned about the reliability of the final point cloud.
• Primary Question: In your experience, can the "increasing time-lag" during playback significantly degrade the spatial accuracy of the scan even if the system doesn't crash?
• Calibration Inquiry: Should I prioritize re-performing the Intrinsic and Extrinsic calibration for all sensors to resolve these accuracy/sync issues? Or is it more likely that the spatial misalignment I see is a direct byproduct of the temporal (timestamp) drift rather than poor calibration parameters?
I have attached screenshots of the generated scans (Photos 3 & 4) for your reference. I would be very grateful for any insights you could provide on whether my focus should be on refining the hardware sync code or on the calibration process.
Thank you for your time and for your contribution to the robotics community.
Best regards,
HJ Kim