TY - GEN
T1 - NRTIRL Based NN-RRT* Path Planner in Human-Robot Interaction Environment
AU - Wang, Yao
AU - Kong, Yuqi
AU - Ding, Zhiyu
AU - Chi, Wenzheng
AU - Sun, Lining
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022.
PY - 2022
Y1 - 2022
N2 - In human-robot interaction environment, it is of great significance for mobile robots to have the awareness of social rules, to realize the socialization and anthropomorphism of robot navigation behavior, and enhance the scene adaptation ability of socialized navigation. Learning from Demonstration (LfD) can obtain an optimized robot trajectory by learning the expert path. Inspired by the LfD method, we propose a new navigation method to learn navigation behaviors form demonstration paths of experts by Neural Network Rapidly-exploring Random Trees (NN-RRT*) planner in the human-robot interaction environment. First, we propose a new NN-RRT* planner to generate paths. Next, the features of demonstration paths and planned paths are extracted for Inverse Reinforcement Learning (IRL) process. The cost function of the path planner is updated. Finally, a new NN-RRT* that can adapt to the complex human-robot interaction environment is obtained. The experimental results show that comparing with the state-of-the-art methods, the path generated by the new navigation method has a higher degree of anthropomorphism and is suitable for navigation in a complex human-robot interaction environment.
AB - In human-robot interaction environment, it is of great significance for mobile robots to have the awareness of social rules, to realize the socialization and anthropomorphism of robot navigation behavior, and enhance the scene adaptation ability of socialized navigation. Learning from Demonstration (LfD) can obtain an optimized robot trajectory by learning the expert path. Inspired by the LfD method, we propose a new navigation method to learn navigation behaviors form demonstration paths of experts by Neural Network Rapidly-exploring Random Trees (NN-RRT*) planner in the human-robot interaction environment. First, we propose a new NN-RRT* planner to generate paths. Next, the features of demonstration paths and planned paths are extracted for Inverse Reinforcement Learning (IRL) process. The cost function of the path planner is updated. Finally, a new NN-RRT* that can adapt to the complex human-robot interaction environment is obtained. The experimental results show that comparing with the state-of-the-art methods, the path generated by the new navigation method has a higher degree of anthropomorphism and is suitable for navigation in a complex human-robot interaction environment.
KW - IRL
KW - NN-RRT
KW - Robot navigation
UR - https://www.scopus.com/pages/publications/85149880029
U2 - 10.1007/978-3-031-24667-8_44
DO - 10.1007/978-3-031-24667-8_44
M3 - 会议稿件
AN - SCOPUS:85149880029
SN - 9783031246661
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 496
EP - 508
BT - Social Robotics - 14th International Conference, ICSR 2022, Proceedings
A2 - Cavallo, Filippo
A2 - Fiorini, Laura
A2 - Sorrentino, Alessandra
A2 - Cabibihan, John-John
A2 - He, Hongsheng
A2 - Liu, Xiaorui
A2 - Matsumoto, Yoshio
A2 - Ge, Shuzhi Sam
PB - Springer Science and Business Media Deutschland GmbH
T2 - 14th International Conference on Social Robotics, ICSR 2022
Y2 - 13 December 2022 through 16 December 2022
ER -