Mobile Robot Exploration Based on Rapidly-exploring Random Trees and Dynamic Window Approach
2019
Exploration is a critical function for autonomous mobile robots. Traditionally, the entire map has to be processed to extract frontiers and perform path planning. However, as the robot explores the environment, the map grows over time, and increasing computational resources are required, especially for large-scale environments. Moreover, only a few methods focus on the exploration on point cloud maps. Here, I present a new practical method to autonomous mobile robot exploration based on a sparse, relatively small-size point cloud local map, which combines Rapidly-exploring Random Tree (RRT) and dynamic window approach (DWA) algorithm together. The local map is built from the consecutive inputs of raw point clouds using an inexpensive 3D sensor, i.e. Kinect V2. Frontiers are effectively detected and local path planning is performed by RRT algorithm directly on unordered point cloud local maps. Motion planning is performed online by DWA to avoid obstacles and direct a nonholonomic mobile robot towards frontiers separating known environments from unknown environments. Embedded with simultaneous localization and mapping (SLAM) system of my previous research, the performance of the proposed method is evaluated in a large-scale customized virtual environment with a size of $33 \times 29 \times 6\mathrm{m}$ using Gazebo as the robotic simulator. The results suggest that the proposed algorithm can accurately direct the nonholonomic mobile robot to unexplored environments in real time. Also, it successfully helps build a coherent semi-metric topological map. The proposed algorithm shows great efficient performance and is suitable for both static and dynamic environments. Combination of RRT and DWA algorithm on point clouds, as a general approach, can be extended to generic 3D nonplanar physical environments. Videos of the experiments can be found at https://youtu.be/0i766fhs9Ds.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
24
References
3
Citations
NaN
KQI