Simultaneous localization and mapping (SLAM) in unstructured environments remains a fundamental challenge for autonomous mobile robots due to sparse geometric features, sensor degradation, and environmental dynamics. In this study, a robust adaptive multi-sensor SLAM framework that tightly integrates lidar, vision, and inertial data via factor graph optimization. A novel adaptive fusion strategy dynamically adjusts the contribution of each sensor based on real-time confidence estimates, while IMU pre-integration and online lidar-vision calibration further enhance motion prediction and cross-modal consistency. Extensive experiments are conducted on benchmark datasets and custom unstructured field data to evaluate the performance of the system. Results show that our approach significantly reduces localization error and improves loop closure accuracy compared to several state-of-the-art SLAM frameworks. Moreover, the system maintains real-time processing capability and low computational overhead, making it suitable for deployment on embedded robotic platforms. This work not only improves the robustness and adaptability of SLAM in complex real-world scenarios, but also lays the foundation for future extensions, including multi-robot collaboration, learning-enhanced perception, and cloud-based large-scale SLAM systems.
Wang et al. (Fri,) studied this question.