日用品形状と人動作の統計的学習に基づくアーム動作生成

抄録

This paper presents a method for generating robot motions for daily tasks by learning the relationship between daily life objects shapes and human motions. Robots are required to handle multifarious objects in various categories. However, it is difficult for robots to plan motions automatically for performing tasks because many objects, including objects in the same category, have different shapes. In the proposed method, object shapes are estimated from a statistical shape model and motions are generated by learning the relationship between object shapes and human motions via linear regression analysis. We applied the proposed method to a task of scooping beans with spoons. We evaluate the generated motions and experimental results of the tasks, which were performed by a robot, to verify the effectiveness of the proposed method.