Improved Trajectory-based LiDAR-INS Hand-Eye Calibration Method

Calibration between different sensors is crucial for improving mapping accuracy. Some sensors, such as inertial navigation systems (INS), cannot extract features and rely on motion-based hand-eye calibration to determine their transformation with other sensors. However, existing methods neglect cumulative errors in light detection and ranging (LiDAR) or camera-based mapping and inertial measurement unit (IMU) drift in rotation, degrading calibration accuracy. In this paper, an improved trajectory-based LiDAR-INS hand-eye calibration method is proposed. Firstly, a novel residual that accounts for IMU rotational error is designed, distributing the error across each trajectory vertex. Then, a weight matrix related to the distances between vertices in the trajectory graph is introduced, assigning higher weights to vertices with loop closure relationships. Finally, the proposed calibration method is validated on our self-collected data sequences. Experiments show that the proposed method outperforms the baseline, with an average loss of 0.0112 per vertex across four sequences. In addition, the method converges in an average of 22 iterations across these sequences, enabling real-time computation.