Prediction-based synchronized human motion imitation by a humanoid robot

In this paper we propose a framework of realtime whole-body human motion imitation for humanoid robots. The approach starts with a kinematic mapping in combination with a balancing algorithm in order to ensure the dynamic constraints during different stance phases. In order to compensate for time delays in the imitation, which emerge from the required weight transfer before stance changes, we apply a machine learning approach based on Hidden Markov Models and Gaussian Regression. Once locomotion primitives are learned from demonstrations, the robot can recognize human's current locomotion state and predict future trajectories. The proposed approaches are implemented and evaluated using a small humanoid robot NAO.