Growing neural gas based navigation system in unknown terrain environment for an autonomous mobile robot

Recently, various types of autonomous robots have been expected in many fields such as a disaster site, forest, and so on. The autonomous robots are assumed to be utilized in unknown environments. In such environments, a path planning to a target point set in the unknown area is a fundamental capability for efficiently executing tasks. To realize the 3D space perception, GNG with Different Topologies (GNG-DT) proposed in our previous work can learn the multiple topological structures with in the framework of learning algorithm. This paper proposes a GNG-DT based 3D perception method by utilizing the multiple topological structures for perceiving the 3D unknown terrain environment and a path planning method to the target point set in the unknown area. Especially, a traversability property of the robot is added to GNG-DT as a new property of the topological structures for clustering the 3D terrain environment from the 3D point cloud measured by 3D Lidar. Furthermore, this paper proposes a path planning method utilizing the multiple topological structures. Next, this paper shows several experimental results of the proposed method using simulation terrain environments for verifying the effectiveness of our proposed method. Finally, we summarize our proposed method and discuss the future direction on this research.

This version of the article has been accepted for publication, after peer review (when applicable) and is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: http://dx.doi.org/10.1007/s10015-022-00826-y

This full-text file will be available in Nov. 2023.