Continuous, reliable, and accurate pose estimation is the basis for perception, planning, and control, which are parts of an autonomous vehicle system or a mobile robot system. For complementarity to each other, the global navigation satellite system (GNSS) can achieve accurate pose solutions in many cases after being integrated with the strap-down inertial navigation system (SINS), which can be further enhanced via fusing visual information. However, unmodeled errors will be introduced in the multi-sensor fusion system by inaccurate spatial parameters caused by coarse calibration or slight changes in mechanical structure. Moreover, an unsynchronized temporal relationship will also influence the pose and extrinsic accuracies. To reduce the negative influence brought by spatial and temporal misalignment, we propose a monocular camera-aided GNSS/SINS integration system with online spatial and temporal compensation, where the time delay and extrinsic of SINS and camera will be estimated and compensated online together with the SINS state. The influence of temporal compensation and GNSS information on extrinsic accuracy is analyzed with a Monte Carlo simulation. The Monte Carlo simulation demonstrates that the extrinsic accuracy improves by 31.33% and 70.00% for rotation and translation after adding temporal compensation and GNSS information. Additionally, the performance of the proposed multi-sensor fusion system is analyzed with two real-world experiments. The position accuracy improves by 19.16%~31.04% in both experiments, which further reveals that online spatial-temporal compensation can significantly improve the pose accuracy of the fusion system.

Visual-Inertial extrinsic calibration is essential for achieving accurate and reliable data fusion in multi-source integrated navigation and positioning systems. However, most traditional methods rely on trajectory optimization driven by image features and inertial pre-integration, which are highly susceptible to noise, fitting errors, and other interfering factors. To address this limitation, we propose a novel visual-inertial calibration framework. Specifically, the framework employs an efficient Global Navigation Satellite System/Inertial Navigation System (GNSS/INS) postprocessing method to generate high-precision carrier trajectories, thereby replacing the poses derived from inertial navigation. Meanwhile, it computes the spatial coordinates of landmarks in the Earth-Centered Earth-Fixed (ECEF) coordinate system and reconstructs the visual projection equation. A unified global least-squares optimization problem is formulated to obtain the globally optimal extrinsic parameters. The Monte Carlo simulations analyze the impact of reference poses on calibration accuracy, demonstrating that our proposed algorithm achieves at least 54.1% and 54.6% improvements in attitude and translation accuracy, respectively, compared with the Kalibr toolbox. Additionally, three real-world experiments evaluate the positioning accuracy of visual-inertial odometry (VIO) to inversely verify the calibration performance. The results demonstrate that, across all datasets, the VIO system calibrated with the proposed extrinsic parameters achieves significant position accuracy improvements of 45.3%, 14.3%, and 30.3%, respectively. This further validates the feasibility and superiority of the GNSS/INS post-processing-assisted extrinsic parameter calibration method.

Dual-LiDAR systems are widely deployed in autonomous driving, yet extrinsic calibration remains challenging in non-overlapping field-of-view (FoV) configurations where correspondence-based methods are unreliable. We propose an engineering-oriented 2.5D calibration framework that estimates horizontal extrinsics $(x,y,\mathrm{yaw})$ via motion-guided planar alignment and then refines them using Gaussian Process Implicit Surfaces (GPIS), which provide continuous and probabilistic surface constraints from spatially disjoint scans. This design avoids calibration targets and reduces dependence on strong scene assumptions, improving robustness under noise and weak structure. Extensive high-fidelity simulation experiments demonstrate centimeter-level lateral accuracy and sub-degree yaw error, consistently outperforming representative motion-based and BEV-based baselines under both clean and noisy settings. To further assess real-world applicability, we conduct a preliminary nuScenes case study by splitting LiDAR scans into front and rear subsets to emulate a non-overlapping dual-LiDAR setup, achieving improved yaw accuracy and competitive lateral precision. Overall, the proposed method serves as a practical refinement stage for non-overlapping dual-LiDAR calibration, with a favorable balance of accuracy, robustness, and engineering feasibility.