Air–Ground Collaborative Autonomous Exploration and Mapping Method for Complex Multi-Grain Pile Environments

Prompt 3D mapping of grain storage is essential for effective management. However, standard mapping algorithms encounter a number of challenges, with the typical granary environment containing dust, grain piles, and narrow aisles. A single robotic agent is not able to provide complete area coverage, and most multi-robot approaches involve re-scanning the same areas due to a lack of explicit viewpoint-based task allocation processes. In order to overcome the above issues, we propose an air–ground collaborative exploration system for complex multi-grain pile scenarios. Exploration redundancy can be reduced by estimating the advantages of viewpoints through ray tracing and assigning the tops of the grain piles to aerial robots with ground vehicles in lower regions and narrow aisles. In order to manage dense dust (5–15 mg/m3), the quality-aware fusion strategy evaluates the reliability of the distance and point density of the sensing to reduce the influence of degraded aerial depth data. Moreover, mapping relies on LiDAR data to ensure mapping quality. A mechanism for re-scanning to enable coverage-driven exploitation of insufficiently explored regions is subsequently proposed. The simulation results show that the design achieved a grain pile coverage of 97.2%, with the total exploration time reduced by 20.1% over single-robot baselines. The results indicate that viewpoint-aware task allocation and dust-sensitive perception fusion can offer a practical solution for autonomous inspection in GPS-restricted, dust-rich industrial environments, such as granary facilities.