Posture imitation and balance learning for humanoid robots

In robotics, controlling the robot by imitating human posture is going to be a tendency. To accomplish the objective of the posture imitation via the robot, there are two main points: extracting useful information from the human motion and solving the falling problem caused by the insufficient sense of balance of the robot. This paper presents a conceptual approach for human posture imitation with balance control on the humanoid robot. First, this paper uses Kinect to capture the human motion and extract key posture by calculating dissimilarity value. Furthermore, the same key postures are clustered via Online Clustering method. Finally, the Balance Learning is used to rebuild the sense of balance of the robot. The experiment results show that the proposed method can achieve the objectives of imitating human postures.