A Vision-Guided Multi-Robot Cooperation Framework for Learning-by-Demonstration and Task Reproduction.