Supplementary

Why FAST Keypoint?

Sensor Compatibility:
- KITTI lacks good IMU data, which is critical for tightly-coupled LIO methods.
Overfitting Concerns:
- Existing algorithms have become over-optimized on datasets like MulRan and Oxford RobotCar, leaving limited room for meaningful comparison or breakthroughs in performance.
Ground Truth Accuracy:
- Large-scale environments in these datasets may have unreliable ground truth, affecting evaluation validity.

Diverse Environments:
- Include varied scenarios and sensor configurations.
Real-World Relevance:
- MCD's ATV and M2DGR's ground robot align with practical robotics use cases.
Robust Evaluation:
- Provide challenging conditions to validate method robustness.