A Social Navigation Framework Based on Laplace Group Dynamics for Humanoid Robot in the Dynamic Scenario

Nowadays, humanoid robots are expected to possess perception and interaction abilities that can be applied in social applications. In this paper, efforts are made to enable a humanoid robot to identify potentially interactive human groups while avoiding pedestrians in dynamic scenarios. To this end, an integrated framework is proposed. In this framework, a group identification module is first developed to classify individuals into groups and effectively infer the group behavioral trend by monitoring their temporal characteristics. Then, an interactive decision maker based on Laplacian dynamics is established to guide the humanoid robot in making interactive decisions to ensure the target groups. Based on the above perception and decision designs, a novel robot behavior planning algorithm is designed to both approach interactive targets in a socially compliant manner and avoid non-interactive ones. Simulation and experiments based on a humanoid robot have demonstrated that the framework can achieve higher social planning speed and better interaction quality while maintaining satisfactory avoidance performance. These findings indicate that the Laplacian-based decision-maker can enable humanoid robots to efficiently respond to groups with various behavior patterns.