This work proposes newly an incremental motion learning algorithm through kinesthetic teachings and a new motion production algorithm by combining learned motions in a humanoid robot. The proposed algorithms are useful to enrich producible motions by a humanoid robot. The learning algorithm consists of time alignment, data integration, dimensional reduction, parameter initialization in the Gaussian mixture model (GMM) of motions, GMM refinement, and motion generation steps. The overall procedure is built to be incremental. No historic data memorization is required in every step and model parameters are enough information to generate motions. The motion production algorithm allows a robot to extract new motions simply from learned motions without requiring teaching sessions. A series of experiments with a Nao humanoid robot validate the performance of the proposed algorithms.
Related publications
1. S Cho, S Jo, Incremental Online Learning of Robot Behaviors from Selected Multiple Kinesthetic Teaching Trials, IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2013. [LINK] [PDF]
2. S Cho, S Jo, Incremental motion learning through kinesthetic teachings and new motion production from learned motions by a humanoid robot, International Journal of Control Automation and Systems (IJCAS), 10(1), 2012. [LINK] [PDF]
3.B Shin, S Jo, Pattern-Preserving-based Motion Imitation for Robots, Proc. of International Conference on Ubiquitous Robots and Ambient Intelligence (URAI), 2011. [PDF]
4. S Cho, S Jo, Kinesthetic learning of behaviors in a humanoid robot, Proc of Int Conf on Control, Automation and Systems (ICCAS) 2011. [PDF]
5. C Hyun, S Jo, Dynamically stable movement generation of a humanoid robot from demonstration: kicking a ball, International Conference on Ubiquitous Robots & Ambient Intelligence (URAI) 2010. [PDF]
