Socially Aware Robot Navigation with Probabilistic Long-Term Human Trajectory Estimation in Dynamic Environments

This paper aims to improve the efficiency of 2D LiDAR-based robot navigation, which can be decreased in dynamic environments. In existing methods, dynamic objects are considered as LiDAR scan data itself, but human behavior patterns (trajectory, speed, etc.) are not considered, which may cause inconvenience to people while the robot is moving along the path. Therefore, in this paper, human behavior patterns (trajectory, speed) are added to a costmap to be considered as obstacles. This enables the robot to perform efficient socially aware robot navigation by considering the human information and avoiding humans in advance. The proposed method is compared with existing methods in a simulation environment by setting the human speed, human trajectory, and the robot’s initial position and goal point under each different condition. The overall results show that the existing methods violated the social distance, while the proposed method does not disturb humans through efficient socially aware robot navigation that considers and avoids humans in advance.