Researchers at Huazhong University of Science and Technology have developed a hierarchical and ordered exploration method for mobile robots that addresses common issues such as repeated exploration and incomplete coverage. The method decouples exploration into two stages: local forward exploration using a soft actor-critic (SAC) algorithm, and global reverse backtracking based on a multi-tree trajectory structure.
Key takeaways
- Local exploration uses a SAC-based policy (LESAC) that takes local occupancy maps as input and combines exploration and safety rewards for efficient depth-first coverage.
- Global backtracking employs a potential field method to identify backtracking points and image masking to optimize paths, ensuring orderly exploration of remaining unknown areas.
- Experiments show higher coverage and efficiency compared to frontier-based and next-best-view methods; ablation studies confirm the necessity of both modules.
Real-machine tests further validated faster exploration speed and higher efficiency, indicating potential for autonomous mobile robots in industrial environments requiring complete mapping.
Source: 《机器人》期刊 (robot.sia.cn) · Published 2026-07-15 · “基于强化学习的移动机器人分层有序环境探索方法”