TY - GEN
T1 - Growing Neural Gas based Traversability Clustering for an Autonomous Robot
AU - Ozasa, Koki
AU - Toda, Yuichiro
AU - Matsuno, Takayuki
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - One of the most important capabilities of an autonomous robot is to recognize a 3D space of the surrounding environment in real-time from a 3D point cloud measured by a 3D distance sensor. The area in which the robot can travel is limited by a robot embodiment such as a mechanism of the robot. Therefore, the traversability estimation method helps the robot to travel safely and reduces the calculation cost of the path planning. This paper proposes Growing Neural Gas (GNG) based traversability estimation method by utilizing a topological structure learned from the 3D point cloud data. However, the conventional GNG cannot preserve the geometric information of the 3D point cloud if the input vector is composed of the multiple properties. Therefore, this paper apply GNG with Different Topologies (GNG-DT) that learn the multiple topological structures according to the number of properties. This paper proposes a GNG-DT based traversability estimation method by redefining the property of the GNG-DT. We conduct several experiments in both simulation and real environment to verify the effectiveness of our proposed method.
AB - One of the most important capabilities of an autonomous robot is to recognize a 3D space of the surrounding environment in real-time from a 3D point cloud measured by a 3D distance sensor. The area in which the robot can travel is limited by a robot embodiment such as a mechanism of the robot. Therefore, the traversability estimation method helps the robot to travel safely and reduces the calculation cost of the path planning. This paper proposes Growing Neural Gas (GNG) based traversability estimation method by utilizing a topological structure learned from the 3D point cloud data. However, the conventional GNG cannot preserve the geometric information of the 3D point cloud if the input vector is composed of the multiple properties. Therefore, this paper apply GNG with Different Topologies (GNG-DT) that learn the multiple topological structures according to the number of properties. This paper proposes a GNG-DT based traversability estimation method by redefining the property of the GNG-DT. We conduct several experiments in both simulation and real environment to verify the effectiveness of our proposed method.
KW - Growing Neural Gas
KW - Traversability estimation
KW - Unsupervised learning
UR - http://www.scopus.com/inward/record.url?scp=85169571400&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85169571400&partnerID=8YFLogxK
U2 - 10.1109/IJCNN54540.2023.10191416
DO - 10.1109/IJCNN54540.2023.10191416
M3 - Conference contribution
AN - SCOPUS:85169571400
T3 - Proceedings of the International Joint Conference on Neural Networks
BT - IJCNN 2023 - International Joint Conference on Neural Networks, Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2023 International Joint Conference on Neural Networks, IJCNN 2023
Y2 - 18 June 2023 through 23 June 2023
ER -