Real-time global action planning for unmanned ground vehicle exploration in Three-dimensional spaces

Autonomous exploration is a technical challenge in the field of mobile robots. In this paper, the problem of autonomous unmanned ground vehicle (UGV) exploration is formulated as a continuous action planning process and a novel method to achieve real-time action planning in 3D spaces is proposed. First, a traversability map and reduced approximated generalized Voronoi graph (RAGVG) are introduced as grid and topological map models, respectively. Then, an efficient and robust algorithm is presented for constructing RAGVGs, and a three-stage approach for fast path search is proposed. Furthermore, a fuzzy decision-making approach is employed to evaluate candidates and create action plans. The proposed method is tested in synthetic and real-world scenarios. The results of static experiments suggest that the RAGVG can achieve almost full coverage with significantly less redundancy (50 % less vertexes and 23 % less edges); moreover, collision-free action plans can be created in real time (less than 330 ms). Furthermore, the results of dynamic experiments indicate that the proposed method achieves effective, efficient, and stable performance with two exploration tasks in 3D spaces with static and moving obstacles, and it saves at least 10 % of the time required for full coverage exploration. The code related to this project is open-source and available at https://gitee.com/xinkaized/my_exploration.