Impact Factor:4.9

DOI number:10.1038/s41598-026-64648-9

Journal:Scientific Reports

Abstract:Autonomous detection in grain silos faces unique challenges due to confined spaces and dynamic obstacles. Conventional path planning methods often fail to simultaneously meet the requirements of navigation efficiency and trajectory smoothness. The study develops a LiDAR-based system, which is characterized as follows: (1) It employs Gmapping with 5 cm resolution grid mapping. (2) It combines a distance-weighted heuristic (w = 1.2) and an enhanced D*lite algorithm for point density optimization. (3) It’s three-layer architecture for real-time sensor data fusion. The system was tested in both simulated and real silo environments (12 m in diameter and 30 m in height). The outcomes showed that the system achieved a success rate of 95.67% (n = 150 tests), which was significantly better than A* (82.3%, p < 0.01) and rapidly-exploring random tree * (RRT*) (88.9%, p < 0.05). Its curvature optimization reduced the mean path curvature to 0.19 (baseline: 0.43) with a processing time of 7.21 ± 1.3 s. Failure cases (4.33%) occurred mainly in the dead zones of screw structures with widths less than 0.8 m. The average path curvature was 0.19 (baseline: 0.43). In summary, the integrated approach proposed in the study effectively solves the dynamic navigation problem in the confined silo space, and provides a technical basis for intelligent grain silo inspection.

Indexed by:Journal paper

Discipline:Engineering

Document Type:J

ISSN No.:2045-2322

Translation or Not:no

Date of Publication:2026-08-27

Included Journals:SCI、EI